ICAART 2019 Abstracts


Area 1 - Artificial Intelligence

Full Papers
Paper Nr: 6
Title:

Analysis and Classification of Voice Pathologies using Glottal Signal Parameters with Recurrent Neural Networks and SVM

Authors:

Leonardo F. Mendoza, Manoela Kohler, Cristian Muñoz, Evelyn S. Batista and Marco A. Pacheco

Abstract: The classification of voice diseases has many applications in health, in diseases treatment, and in the design of new medical equipment for helping doctors in diagnosing pathologies related to the voice. This work uses the parameters of the glottal signal to help the identification of two types of voice disorders related to the pathologies of the vocal folds: nodule and unilateral paralysis. The parameters of the glottal signal are obtained through a known inverse filtering method and they are used as inputs to an Artificial Neural Network, RNN, LSTM, a Support Vector Machine and also to a Hidden Markov Model, to obtain the classification, and to compare the results, of the voice signals into three different groups: speakers with nodule in the vocal folds; speakers with unilateral paralysis of the vocal folds; and speakers with normal voices, that is, without nodule or unilateral paralysis present in the vocal folds. The database is composed of 248 voice recordings (signals of vowels production) containing samples corresponding to the three groups mentioned. In this study a larger database was used for the classification when compared with similar studies, and its classification rate is superior to other studies, reaching 99.2%.
Download

Paper Nr: 9
Title:

Solving the Social Golfers Problems by Constraint Programming in Sequential and Parallel

Authors:

Ke Liu, Sven Löffler and Petra Hofstedt

Abstract: The social golfer problem (SGP) has received plenty of attention in constraint satisfaction problem (CSP) research as a standard benchmark for symmetry breaking. However, the constraint satisfaction approach has stagnated for solving larger SGP instances over the last decade. We improve the existing model of the SGP by introducing more constraints that effectively reduce the search space, particularly for instances of special form. Furthermore, we present a search space splitting method to solve the SGP in parallel through data-level parallelism. Our implementation of the presented techniques allows us to attain solutions for eight instances with maximized weeks, in which six of them were open instances for the constraint satisfaction approach, and two of them are computed for the first time. Besides, super-linear speedups are observed for all the instances solved in parallel.
Download

Paper Nr: 17
Title:

Cost Partitioning for Multi-agent Planning

Authors:

Michal Štolba, Michaela Urbanovská, Daniel Fišer and Antonín Komenda

Abstract: Similarly to classical planning, heuristics play a crucial role in Multi-Agent Planning (MAP). Especially, the question of how to compute a distributed heuristic so that the information is shared effectively has been studied widely. This question becomes even more intriguing if we aim to preserve some degree of privacy, or admissibility of the heuristic. The works published so far aimed mostly at providing an ad-hoc distribution protocol for a particular heuristic. In this work, we propose a general framework for distributing heuristic computation based on the technique of cost partitioning. This allows the agents to compute their heuristic values separately and the global heuristic value as an admissible sum. We evaluate the presented techniques in comparison to the baseline of locally computed heuristics and show that the approach based on cost partitioning improves the heuristic quality over the baseline.
Download

Paper Nr: 40
Title:

Wide and Deep Reinforcement Learning for Grid-based Action Games

Authors:

Juan M. Montoya and Christian Borgelt

Abstract: For the last decade Deep Reinforcement Learning has undergone exponential development; however, less has been done to integrate linear methods into it. Our Wide and Deep Reinforcement Learning framework provides a tool that combines linear and non-linear methods into one. For practical implementations, our framework can help integrate expert knowledge while improving the performance of existing Deep Reinforcement Learning algorithms. Our research aims to generate a simple practical framework to extend such algorithms. To test this framework we develop an extension of the popular Deep Q-Networks algorithm, which we name Wide Deep Q-Networks. We analyze its performance compared to Deep Q-Networks and Linear Agents, as well as human players. We apply our new algorithm to Berkley’s Pac-Man environment. Our algorithm considerably outperforms Deep Q-Networks’ both in terms of learning speed and ultimate performance showing its potential for boosting existing algorithms.
Download

Paper Nr: 41
Title:

Improving Readability for Tweet Contextualization using Bipartite Graphs

Authors:

Amira Dhokar, Lobna Hlaoua and Lotfi Ben Romdhane

Abstract: Tweet contextualization (TC) is a new issue that aims to answer questions of the form What is this tweet about? The idea of this task was imagined as an extension of a previous area called multi-document summarization (MDS), which consists in generating a summary from many sources. In both TC and MDS, the summary should ideally contain most relevant information of the topic that is being discussed in the source texts (for MDS) and related to the query (for TC). Furthermore of being informative, a summary should be coherent, i.e. well written to be readable and grammatically compact. Hence, coherence is an essential characteristic in order to produce comprehensible texts. In this paper, we propose a new approach to improve readability and coherence for tweet contextualization based on bipartite graphs. The main idea of our proposed method is to reorder sentences in a given paragraph by combining most expressive words detection and HITS (Hyperlink- Induced Topic Search) algorithm to make up a coherent context.
Download

Paper Nr: 49
Title:

The Impact of Environmental and Evolutionary Factors on the Emergence of Cooperation among Evolved Mobile Agents

Authors:

Maud D. Gibbons, Josephine Griffith and Colm O’Riordan

Abstract: This paper presents work investigating the influence of various environmental and evolutionary factors on the evolution of cooperation in a spatial game theoretical setting. These include agent mobility, population density, agent lifespan, and the placement mechanism. In the model considered, a population of agents inhabit a toroidal lattice grid, in which they participate in the Prisoner’s Dilemma game. The agents have the ability to respond to, and learn from, environmental stimuli. In particular, agents learn movement strategies to compete with other agents in the game, which may result in improved payoffs by increasing the number of beneficial interactions. We compare the levels of cooperation and the corresponding movement strategies evolved under the various environmental and evolutionary settings. We present results indicating that, given suitable densities and evolutionary settings, cooperators in well-mixed populations develop a suitable movement strategy to promote the evolution of cooperation. Additionally, we show that cooperation may emerge without significant aid from mobile strategies given a placement mechanism conducive to the formation of cooperator clusters.
Download

Paper Nr: 55
Title:

Reinforcement Learning Approach for Cooperative Control of Multi-Agent Systems

Authors:

Valeria Javalera-Rincon, Vicenc P. Cayuela, Bernardo M. Seix and Fernando Orduña-Cabrera

Abstract: Reinforcement Learning (RL) systems are trial-and-error learners. This feature altogether with delayed reward, makes RL flexible, powerful and widely accepted. However, RL could not be suitable for control of critical systems where the learning of the control actions by trial and error is not an option. In the RL literature, the use of simulated experience generated by a model is called planning. In this paper, the planningByInstruction and planningByExploration techniques are introduced, implemented and compared to coordinate, a heterogeneous multi-agent architecture for distributed Large Scale Systems (LSS). This architecture was proposed by (Javalera 2016). The models used in this approach are part of a distributed architecture of agents. These models are used to simulate the behavior of the system when some coordinated actions are applied. This experience is learned by the so-called, LINKER agents, during an off-line training. An exploitation algorithm is used online, to coordinate and optimize the value of overlapping control variables of the agents in the distributed architecture in a cooperative way. This paper also presents a technique that offers a solution to the problem of the number of learning steps required to converge toward an optimal (or can be sub-optimal) policy for distributed control systems. An example is used to illustrate the proposed approach, showing exciting and promising results regarding the applicability to real systems.
Download

Paper Nr: 62
Title:

Instance-incremental Classification of Imbalanced Bidding Fraud Data

Authors:

Ahmad Alzahrani and Samira Sadaoui

Abstract: Shill Bidding (SB) has been recognized as the predominant online auction fraud and also the most difficult to detect due to its similarity to normal bidding behavior. Previously, we produced a high-quality SB dataset based on actual auctions and effectively labeled the instances into normal or suspicious. To overcome the serious problem of imbalanced SB datasets, in this study, we investigate over- and under-sampling techniques through several instance-based classification algorithms. Thousands of auctions occur in eBay every day, and auction data may be sent continuously to the optimal fraud classifier to detect potential SB activities. Consequently, instance-based classification is appropriate for our particular fraud detection problem. According to the experimental results, incremental classification returns high performance for both over- and under-sampled SB datasets. Still, over-sampling slightly outperforms under-sampling for both normal and suspicious classes across all the classifiers.

Paper Nr: 63
Title:

On the Role of Central Individuals in Influence Propagation

Authors:

Rafael de Santiago, Fernando Concatto and Luís C. Lamb

Abstract: Recently, the influence of individuals in complex networks received the attention of several fields of science. In the context of influence spreading, the understanding of the role and importance of each individual can be used to control the spread of memes. By considering centrality measures as defining factors of individual importance, this paper investigates the relationship between the importance of an individual and its role in the propagation of influence within and over a network. In order to do so, we used degree measures, betweenness centrality, closeness centrality, eigenvector centrality and clustering coefficient over four different real graphs. The Min-SEIS-Cluster model was employed in order to simulate the spread of memes, which involve cutting connections to minimize an epidemic. The results revealed a high correlation between individual importance and prominence on influence propagation, and the potential to utilize centrality measures to identify which connections should be cut off in specific application scenarios.
Download

Paper Nr: 64
Title:

LSTM Neural Networks for Transfer Learning in Online Moderation of Abuse Context

Authors:

Avi Bleiweiss

Abstract: Recently, the impact of offensive language and derogatory speech to online discourse, motivated social media platforms to research effective moderation tools that safeguard internet access. However, automatically distilling and flagging inappropriate conversations for abuse remains a difficult and time consuming task. In this work, we propose an LSTM based neural model that transfers learning from a platform domain with a relatively large dataset to a domain much resource constraint, and improves the target performance of classifying toxic comments. Our model is pretrained on personal attack comments retrieved from a subset of discussions on Wikipedia, and tested to identify hate speech on annotated Twitter tweets. We achieved an F1 measure of 0.77, approaching performance of the in-domain model and outperforming out-domain baseline by about nine percentage points, without counseling the provided labels.
Download

Paper Nr: 66
Title:

Deep Reinforcement Learning for Pellet Eating in Agar.io

Authors:

Nil S. Ansó, Anton O. Wiehe, Madalina M. Drugan and Marco A. Wiering

Abstract: The online game Agar.io has become massively popular on the internet due to its intuitive game design and its ability to instantly match players with others around the world. The game has a continuous input and action space and allows diverse agents with complex strategies to compete against each other. In this paper we focus on the pellet eating task in the game, in which an agent has to learn to optimize its navigation strategy to grow maximally in size within a specific time period. This work first investigates how different state representations affect the learning process of a Q-learning algorithm combined with artificial neural networks which are used for representing the Q-function. The representations examined range from raw pixel values to extracted handcrafted feature vision grids. Secondly, the effects of using different resolutions for the representations are examined. Finally, we compare the performance of different value function network architectures. The architectures examined are two convolutional Deep Q-networks (DQN) of varying depth and one multilayer perceptron. The results show that the use of handcrafted feature vision grids significantly outperforms the direct use of raw pixel input. Furthermore, lower resolutions of 42×42 lead to better performances than larger resolutions of 84 × 84.
Download

Paper Nr: 72
Title:

Hyperparameter Optimization for Deep Reinforcement Learning in Vehicle Energy Management

Authors:

Roman Liessner, Jakob Schmitt, Ansgar Dietermann and Bernard Bäker

Abstract: Reinforcement Learning is a framework for algorithms that learn by interacting with an unknown environment. In recent years, combining this approach with deep learning has led to major advances in various fields. Numerous hyperparameters – e.g. the learning rate – influence the learning process and are usually determined by testing some variations. This selection strongly influences the learning result and requires a lot of time and experience. The automation of this process has the potential to make Deep Reinforcement Learning available to a wider audience and to achieve superior results. This paper presents a model-based hyperparameter optimization of the Deep Deterministic Policy Gradients (DDPG) algorithm and demonstrates it with a hybrid vehicle energy management environment. In the given case, the hyperparameter optimization is able to double the gained reward value of the DDPG agent.
Download

Paper Nr: 76
Title:

Estimation of the Cardiac Pulse from Facial Video in Realistic Conditions

Authors:

Arvind Subramaniam and K. Rajitha

Abstract: Remote detection of the cardiac pulse has a number of applications in fields of sports and medicine, and can be used to determine an individual’s physiological state. Over the years, several papers have proposed a number of approaches to extract heart rate (HR) using video imaging. However, all these approaches have employed the Viola-Jones algorithm for face detection. Additionally, these methods usually require the subject to be stationary and do not take illumination changes into account. The present research proposes a novel framework that employs Faster RCNNs (Region-based Convolutional Neural Networks) for face detection, followed by face tracking using the Kanade-Lukas-Tomasi (KLT) algorithm. In addition, the present framework recovers the feature points which are lost during extreme head movements of the subject. Our method is robust to extreme motion interferences (head movements) and utilizes Recursive Least Square (RLS) adaptive filtering methods to tackle interferences caused by illumination variations. The accuracy of the model has been tested based on a movie evaluation scenario and the accuracy was estimated on a public database MAHNOB-HCI. The output of the performance measure showed that the present model outperforms previously proposed methods.
Download

Paper Nr: 77
Title:

UAV Flocks Forming for Crowded Flight Environments

Authors:

Rina Azoulay and Shulamit Reches

Abstract: In this study, we consider a situation where several privately owned unmanned aerial vehicles (UAVs) are supposed to travel on several routes. We develop a model for grouping them into UAV flocks that are supposed to travel on similar routes within the same time window. Our proposed flocking protocol enables each UAV to optimize its own preferences concerning its flights. Using this protocol enables all UAVs to enjoy freer routes and fewer encounters with other UAVs, thus saving time and energy during their flights. The protocol allows each UAV to create a flock or to join an existing flock to save its resources. Joining flocks in a crowded environment can reduce the overhead caused by encountering additional UAVs in the environment. We developed a flocking protocol that allows each UAV to design its optimal route. The protocol is based on a public on-line communication blackboard, which enables each UAV to receive information about existing flocks, join an existing flock or build a new flock and publish it on the blackboard. In addition, we defined a strategy for each UAV to assist in deciding which flock to join or whether to create a new flock to optimize its expected utility. Finally, the effectiveness of the proposed algorithm is verified by means of simulations.
Download

Paper Nr: 79
Title:

Detecting Adversarial Examples in Deep Neural Networks using Normalizing Filters

Authors:

Shuangchi Gu, Ping Yi, Ting Zhu, Yao Yao and Wei Wang

Abstract: Deep neural networks are vulnerable to adversarial examples which are inputs modified with unnoticeable but malicious perturbations. Most defending methods only focus on tuning the DNN itself, but we propose a novel defending method which modifies the input data to detect the adversarial examples. We establish a detection framework based on normalizing filters that can partially erase those perturbations by smoothing the input image or depth reduction work. The framework gives the decision by comparing the classification results of original input and multiple normalized inputs. Using several combinations of gaussian blur filter, median blur filter and depth reduction filter, the evaluation results reaches a high detection rate and achieves partial restoration work of adversarial examples in MNIST dataset. The whole detection framework is a low-cost highly extensible strategy in DNN defending works.
Download

Paper Nr: 81
Title:

Improving Transfer Learning Performance: An Application in the Classification of Remote Sensing Data

Authors:

Gabriel L. Tenorio, Cristian M. Villalobos, Leonardo F. Mendoza, Eduardo Costa da Silva and Wouter Caarls

Abstract: The present paper aims to train and analyze Convolutional Neural Networks (CNN or ConvNets) capable of classifying plant species of a certain region for applications in an environmental monitoring system. In order to achieve this for a limited training dataset, the samples were expanded with the use of a data generator algorithm. Next, transfer learning and fine tuning methods were applied with pre-trained networks. With the purpose of choosing the best layers to be transferred, a statistical dispersion method was proposed. Through a distributed training method, the training speed and performance for the CNN in CPUs was improved. After tuning the parameters of interest in the resulting network by the cross-validation method, the learning capacity of the network was verified. The obtained results indicate an accuracy of about 97%, which was acquired transferring the pre-trained first seven convolutional layers of the VGG-16 network to a new sixteen-layer convolutional network in which the final training was performed. This represents an improvement over the state of the art, which had an accuracy of 91% on the same dataset.
Download

Paper Nr: 82
Title:

Spatial Kernel Discriminant Analysis: Applied for Hyperspectral Image Classification

Authors:

Soumia Boumeddane, Leila Hamdad, Sophie Dabo-Niang and Hamid Haddadou

Abstract: Classical data mining models relying upon the assumption that observations are independent, are not suitable for spatial data, since they fail to capture the spatial autocorrelation. In this paper, we propose a new supervised classification algorithm which takes into account the spatial dependency of data, named Spatial Kernel Discriminant Analysis (SKDA). We present a non-parametric classifier based on a kernel estimate of the spatial probability density function which combines two kernels: one controls the observed values while the other controls the spatial locations of observations. We applied our algorithm for hyperspectral image (HSI) classification, a challenging task due to the high dimensionality of data and the limited number of training samples. Using our algorithm, the spatial and spectral information of each pixel are jointly used to achieve the classification. To evaluate the efficiency of the proposed method, experiments on real remotely sensed images are conducted, and show that our method is competitive and achieves higher classification accuracy compared to other contextual classification methods.
Download

Paper Nr: 85
Title:

Conflict Handling Framework in Generalized Multi-agent Path finding: Advantages and Shortcomings of Satisfiability Modulo Approach

Authors:

Pavel Surynek

Abstract: We address conflict reasoning in generalizations of multi-agent path finding (MAPF). We assume items placed in vertices of an undirected graph with at most one item per vertex. Items can be relocated across edges while various constraints depending on the concrete type of MAPF must be satisfied. We recall a general problem formulation that encompasses known types of item relocation problems such as multi-agent path finding (MAPF) and token swapping (TSWAP). We show how to express new types of relocation problems in the general problem formulation. We thoroughly evaluate a novel solving method for item relocation that combines satisfiability modulo theory (SMT) with conflict-based search (CBS). CBS is interpreted in the SMT framework where we start with the basic model and refine the model with a collision resolution constraint whenever a collision between items occurs. The key difference between the standard CBS and the SMT-based modification of CBS (SMT-CBS) is that the standard CBS branches the search to resolve the collision while SMT-CBS iteratively adds a single disjunctive collision resolution constraint. Our experimental evaluation revealed that although SMT-CBS performs better than CBS in small densely occupied instances of variants of MAPF, it is outperformed on large sparsely occupied environments. The performed analysis shows that individual paths in large environments of relocation instances can be found faster using simple A*-based algorithm than by the SMT solver. On the other hand the SMT solver performs better when many conflicts between items need to be resolved.
Download

Paper Nr: 91
Title:

A Hybrid Neural Network and Hidden Markov Model for Time-aware Recommender Systems

Authors:

Hanxuan Chen and Zuoquan Lin

Abstract: In this paper, we propose a hybrid model that combines neural network and hidden Markov model for time-aware recommender systems. We use higher-order hidden Markov model to capture the temporal information of users and items in collaborative filtering systems. Because the computation of the transition matrix of higher-order hidden Markov model is hard, we compute the transition matrix by deep neural networks. We implement the algorithms of the hybrid model for offline batch-learning and online updating respectively. Experiments on real datasets demonstrate that the hybrid model has improvement performances over the existing recommender systems.
Download

Paper Nr: 94
Title:

Multi-agent Systems for Estimating Missing Information in Smart Cities

Authors:

Davide A. Guastella, Valérie Camps and Marie-Pierre Gleizes

Abstract: Smart cities aim at improving the quality of life of citizens. To do this, numerous ad-hoc sensors need to be deployed in a smart city to monitor the environmental state. Even if nowadays sensors are becoming more and more cheap their installation and maintenance costs increase rapidly with their number. This paper makes an inventory of the dimensions required for designing an intelligent system to support smart city initiatives. Then we propose a multi-agent based solution that uses a limited number of sensors to estimate at runtime missing information in smart cities using a limited number of sensors.
Download

Paper Nr: 96
Title:

Exhaustive Testing of Trader-agents in Realistically Dynamic Continuous Double Auction Markets: AA Does Not Dominate

Authors:

Dave Cliff

Abstract: We analyse results from over 3.4million detailed market-trading simulation sessions which collectively confirm an unexpected result: in markets with dynamically varying supply and demand, the best-performing automated adaptive auction-market trading-agent currently known in the AI/Agents literature, i.e. Vytelingum’s Adaptive-Aggressive (AA) strategy, can be routinely out-performed by simpler trading strategies. AA is the most recent in a series of AI trading-agent strategies proposed by various researchers over the past twenty years: research papers contributing major steps in this evolution of strategies have been published at IJCAI, in the Artificial Intelligence journal, and at AAMAS. The innovative step taken here is to brute-force exhaustively evaluate AA in market environments that are in various ways more realistic, closer to real-world financial markets, than the simple constrained abstract experimental evaluations routinely used in the prior academic AI/Agents research literature. We conclude that AA can indeed appear dominant when tested only against other AI-based trading agents in the highly simplified market scenarios that have become the methodological norm in the trading-agents academic research literature, but much of that success seems to be because AA was designed with exactly those simplified experimental markets in mind. As soon as we put AA in scenarios closer to real-world markets, modify it to fit those markets accordingly, and exhaustively test it against simpler trading agents, AA’s dominance simply disappears.
Download

Paper Nr: 108
Title:

Pre-indexing Techniques in Arabic Information Retrieval

Authors:

Souheila Ben Guirat, Ibrahim Bounhas and Yahia Slimani

Abstract: Arabic document indexing is yet challenging given the morphological specificities of this language. Although there has been much effort in the field, developing more efficient indexing approaches is more and more demanding. One of the most important issues concerns the choice of the indexing units (e.g. stems, roots, lemmas, etc.) which both enhances retrieval efficiency and optimizes the indexing process. The question is how to process Arabic texts to retrieve the basic forms which better reflect the meaning of words and documents? In the literature several indexing units have been compared, while combining multiple indexes seems to be promising. In our previous works, we showed that hybrid indexes based on stems, patterns and roots enhances results. However, we need to find the optimal weight of each indexing unit. Therefore, this paper proposes to contribute in optimizing hybrid indexing. We compare and evaluate four pre-indexing methods.
Download

Paper Nr: 110
Title:

The Dynamics of Narrow-minded Belief

Authors:

Shoshin Nomura, Norihiro Arai and Satoshi Tojo

Abstract: The purpose of this paper is to consider and formalize an important factor of human intelligence, belief affected by passion, which we call narrow-minded belief. Based on Public Announcement Logic, we define our logic, Logic Of Narrow-minded belief (LON), as that which includes such belief. Semantics for LON is provided by the Kripke-style semantics, and a proof system for it is given by the Hilbert-style proof system. We then provide a proof of the semantic completeness theorem for the proof system with the innermost strategy of reducing a formula for LON. Using LON, we formally analyze the mental state of the hero of Shakespeare’s tragedy Othello as an example of narrow-minded belief and its formalization.
Download

Paper Nr: 113
Title:

A Reinforcement Learning Method to Select Ad Networks in Waterfall Strategy

Authors:

Reza R. Afshar, Yingqian Zhang, Murat Firat and Uzay Kaymak

Abstract: A high percentage of online advertising is currently performed through real time bidding. Impressions are generated once a user visits the websites containing empty ad slots, which are subsequently sold in an online ad exchange market. Nowadays, one of the most important sources of income for publishers who own websites is through online advertising. From a publisher’s point of view it is critical to send its impressions to most profitable ad networks and to fill its ad slots quickly in order to increase their revenue. In this paper we present a method for helping publishers to decide which ad networks to use for each available impression. Our proposed method uses reinforcement learning with initial state-action values obtained from a prediction model to find the best ordering of ad networks in the waterfall fashion. We show that this method increases the expected revenue of the publisher.
Download

Paper Nr: 114
Title:

Liquidity Stress Detection in the European Banking Sector

Authors:

Richard Heuver and Ron Triepels

Abstract: Liquidity stress constitutes an ongoing threat to financial stability in the banking sector. A bank that manages its liquidity inadequately might find itself unable to meet its payment obligations. These liquidity issues, in turn, can negatively impact the liquidity position of many other banks due to contagion effects. For this reason, central banks carefully monitor the payment activities of banks in financial market infrastructures and try to detect early-warning signs of liquidity stress. In this paper, we investigate whether this monitoring task can be performed by supervised machine learning. We construct probabilistic classifiers that estimate the probability that a bank faces liquidity stress. The classifiers are trained on a dataset consisting of various payment features of European banks and which spans several known stress events. Our experimental results show that the classifiers detect the periods in which the banks faced liquidity stress reasonably well.
Download

Paper Nr: 118
Title:

A Hybrid Intelligent Agent for Notification of Users Distracted by Mobile Phones in an Urban Environment

Authors:

Thiago Â. Gelaim, Gabriel A. Langer, Elder R. Santos, Ricardo A. Silveira, John O’Hare, Paul Kendrick and Bruno M. Fazenda

Abstract: Mobile devices are now ubiquitous in daily life and the number of activities that can be performed using them is continually growing. This implies increased attention being placed on the device and diverted away from events taking place in the surrounding environment. The impact of using a smartphone on pedestrians in the vicinity of urban traffic has been investigated in a multimodal, fully immersive, virtual reality environment. Based on experimental data collected, an agent to improve the attention of users in such situations has been developed. The proposed agent uses explicit, contextual data from experimental conditions to feed a statistical learning model. The agent’s decision process is aimed at notifying users when they become unaware of critical events in their surroundings.
Download

Paper Nr: 120
Title:

Determining Capacity of Shunting Yards by Combining Graph Classification with Local Search

Authors:

Arno van de Ven, Yingqian Zhang, Wan-Jui Lee, Rik Eshuis and Anna Wilbik

Abstract: Dutch Railways (NS) uses a shunt plan simulator to determine capacities of shunting yards. Central to this simulator is a local search heuristic. Solving this capacity determination problem is very time consuming, as it requires to solve an NP-hard shunting planning problem, and furthermore, the capacity has to determined for a large number of possible scenarios at over 30 shunting yards in The Netherlands. In this paper, we propose to combine machine learning with local search in order to speed up finding shunting plans in the capacity determination problem. The local search heuristic models the activities that take place on the shunting yard as nodes in an activity graph with precedence relations. Consequently, we apply the Deep Graph Convolutional Neural Network, which is a graph classification method, to predict whether local search will find a feasible shunt plan given an initial solution. Our experimental results show our approach can significantly reduce the simulation time in determining the capacity of a given shunting yard. This study demonstrates how machine learning can be used to boost optimization algorithms in an industrial application.
Download

Paper Nr: 125
Title:

Pareto-based Soft Arc Consistency for Multi-objective Valued CSPs

Authors:

Limeme Ben Ali, Maher Helaoui and Wady Naanaa

Abstract: A valued constraint satisfaction problem (VCSP) is a soft constraint framework that can formalize a wide range of applications related to Combinatorial Optimization and Artificial Intelligence. Most researchers have focused on the development of algorithms for solving mono-objective problems. However, many real-world satisfaction/optimization problems involve multiple objectives that should be considered separately and satisfied/optimized simultaneously. Solving a Multi-Objective Optimization Problem (MOP) consists of finding the set of all non-dominated solutions, known as the Pareto Front. In this paper, we introduce multi-objective valued constraint satisfaction problem (MO-VCSP), that is a VCSP involving multiple objectives, and we extend soft local arc consistency methods, which are widely used in solving Mono-Objective VCSP, in order to deal with the multi-objective case. Also, we present multi-objective enforcing algorithms of such soft local arc consistencies taking into account the Pareto principle. The new Pareto-based soft arc consistency (P-SAC) algorithms compute a Lower Bound Set of the efficient frontier. As a consequence, P-SAC can be integrated into a Multi-Objective Branch and Bound (MO-BnB) algorithm in order to ensure its pruning efficiency.
Download

Paper Nr: 129
Title:

Spectral Algorithm for Line Graphs to Find Overlapping Communities in Social Networks

Authors:

Camila S. Tautenhain and Mariá V. Nascimento

Abstract: A great deal of community detection communities is based on the maximization of the measure known as modularity. There is a dearth of literature on overlapping community detection algorithms, in spite of the importance of the applications and the overwhelming number of community detection algorithms yet proposed. To this end, one of the suggestions in the literature consists of partitioning the set of edges into communities, also known as link partitions, by applying community detection algorithms to line graphs. In line with this, in this paper, overlapping vertex communities are obtained from link partitions by a method that selects the communities of the edges that represent the highest modularity gain. We also introduce a spectral algorithm to find link partitions from line graphs. We show that the modularity of communities in line graphs is equivalent to the adaptation of modularity of communities in the original graphs, when considering the non-backtracking matrix instead of the adjacency matrix in its formula. The results of the experiments carried out with overlapping community detection algorithms showed that the proposed method is competitive with state-of-the-art algorithms.
Download

Paper Nr: 131
Title:

Hierarchical Reinforcement Learning Introducing Genetic Algorithm for POMDPs Environments

Authors:

Kohei Suzuki and Shohei Kato

Abstract: Perceptual aliasing is one of the major problems in applying reinforcement learning to the real world. Perceptual aliasing occurs in the POMDPs environment, where agents cannot observe states correctly, which makes reinforcement learning unsuccessful. HQ-learning is cited as a solution to perceptual aliasing. HQ-learning solves perceptual aliasing by using subgoals and subagent. However, subagents learn independently and have to relearn each time when subgoals change. In addition, the number of subgoals is fixed, and the number of episodes in reinforcement learning increases unless the number of subgoals is appropriate. In this paper, we propose the reinforcement learning method that generates subgoals using genetic algorithm. We also report the effectiveness of our method by some experiments with partially observable mazes.
Download

Paper Nr: 136
Title:

A PSO based Approach to Assign Segments for Reducing Excavated Soil in Shield Tunneling

Authors:

Koya Ihara, Shohei Kato, Takehiko Nakaya, Tomoaki Ogi and Hiroichi Masuda

Abstract: It is expected that artificial intelligence reduce labor and improve productivity of the shield tunneling, which is one of the tunnel construction method. In the planning process of the shield tunneling, segments of the tunnel are assigned along to the predetermined curve called the planning line. Conventionally, skilled engineers manually assign the segments to minimize the amount of gaps between each segment and the planning line. Nevertheless, we have only to reduce each gap less than a tolerance, and there is a demand to reduce the amount of soil excavated along to the segments. Handling the reducing gaps as constraints and reducing the amount of excavated soil as an objective, this paper addresses the segment assignment as a constrained combinatorial optimization problem. These constraints are severe, and the problem have an extremely narrow feasible region. To solve the problem we proposed e constrained discrete genetic algorithm (eDGA) and e constrained integer categorical particle swarm optimization (eICPSO), adapting a constraint handling method called the e constrained method to the discrete genetic algorithm and the integer categorical particle swarm optimization. The effectiveness of the eDGA and eICPSO to the segment assignment is shown by the two-dimensional simulator experiment using real construction data. The experimental results show that the proposed method have a potential to find the segment assignment reducing the amount of excavated soil as compared to the conventional method (skilled engineer) while keeping the all gaps between segments and the planning line falling within the tolerance.
Download

Paper Nr: 169
Title:

A Split-Merge Evolutionary Clustering Algorithm

Authors:

Veselka Boeva, Milena Angelova and Elena Tsiporkova

Abstract: In this article we propose a bipartite correlation clustering technique that can be used to adapt the existing clustering solution to a clustering of newly collected data elements. The proposed technique is supposed to provide the flexibility to compute clusters on a new portion of data collected over a defined time period and to update the existing clustering solution by the computed new one. Such an updating clustering should better reflect the current characteristics of the data by being able to examine clusters occurring in the considered time period and eventually capture interesting trends in the area. For example, some clusters will be updated by merging with ones from newly constructed clustering while others will be transformed by splitting their elements among several new clusters. The proposed clustering algorithm, entitled Split-Merge Evolutionary Clustering, is evaluated and compared to another bipartite correlation clustering technique (PivotBiCluster) on two different case studies: expertise retrieval and patient profiling in healthcare.
Download

Paper Nr: 171
Title:

Exploring the Context of Recurrent Neural Network based Conversational Agents

Authors:

Raffaele Piccini and Gerasimos Spanakis

Abstract: Conversational agents have begun to rise both in the academic (in terms of research) and commercial (in terms of applications) world. This paper investigates the task of building a non-goal driven conversational agent, using neural network generative models and analyzes how the conversation context is handled. It compares a simpler Encoder-Decoder with a Hierarchical Recurrent Encoder-Decoder architecture, which includes an additional module to model the context of the conversation using previous utterances information. We found that the hierarchical model was able to extract relevant context information and include them in the generation of the output. However, it performed worse (35-40%) than the simple Encoder-Decoder model regarding both grammatically correct output and meaningful response. Despite these results, experiments demonstrate how conversations about similar topics appear close to each other in the context space due to the increased frequency of specific topic-related words, thus leaving promising directions for future research and how the context of a conversation can be exploited.
Download

Paper Nr: 186
Title:

Multi-Label Network Classification via Weighted Personalized Factorizations

Authors:

Ahmed Rashed, Josif Grabocka and Lars Schmidt-Thieme

Abstract: Multi-label network classification is a well-known task that is being used in a wide variety of web-based and non-web-based domains. It can be formalized as a multi-relational learning task for predicting nodes labels based on their relations within the network. In sparse networks, this prediction task can be very challenging when only implicit feedback information is available such as in predicting user interests in social networks. Current approaches rely on learning per-node latent representations by utilizing the network structure, however, implicit feedback relations are naturally sparse and contain only positive observed feedbacks which mean that these approaches will treat all observed relations as equally important. This is not necessarily the case in real-world scenarios as implicit relations might have semantic weights which reflect the strength of those relations. If those weights can be approximated, the models can be trained to differentiate between strong and weak relations. In this paper, we propose a weighted personalized two-stage multi-relational matrix factorization model with Bayesian personalized ranking loss for network classification that utilizes basic transitive node similarity function for weighting implicit feedback relations. Experiments show that the proposed model significantly outperforms the state-of-art models on three different real-world web-based datasets and a biology-based dataset.
Download

Paper Nr: 191
Title:

Continual Representation Learning for Images with Variational Continual Auto-Encoder

Authors:

Ik H. Jeon and Soo Y. Shin

Abstract: We propose a novel architecture for the continual representation learning for images, called variational continual auto-encoder (VCAE). Our approach builds a time-variant parametric model that generates images close to the observation by using optimized approximate inference over time. When the dataset is sequentially observed, the model efficiently learns underlying representations without forgetting previously acquired knowledge. Through experiments, we evaluate the development of test log-likelihood over time, which shows resistance to the catastrophic forgetting. The results show that VCAE has stronger immunity against catastrophic forgetting in comparison to the benchmark while VCAE requires much less time for training.
Download

Short Papers
Paper Nr: 7
Title:

An Extended Paradefinte Belnap–Dunn Logic that is Embeddable into Classical Logic and Vice Versa

Authors:

Norihiro Kamide

Abstract: In this study, an extended paradefinite Belnap–Dunn logic (PBD) is introduced as a Gentzen-type sequent calculus. The logic PBD is an extension of Belnap–Dunn logic as well as a modified subsystem of Arieli, Avron, and Zamansky’s ideal four-valued paradefinite logic known as 4CC. The logic PBD is formalized on the basis of the idea of De and Omori’s characteristic axiom scheme for an extended Belnap–Dunn logic with classical negation (BD+), even though PBD has no classical negation connective but can simulate classical negation. Theorems for syntactically and semantically embedding PBD into a Gentzen-type sequent calculus for classical logic and vice versa are proved. The cut-elimination and completeness theorems for PBD are obtained via these embedding theorems.
Download

Paper Nr: 14
Title:

Chaos-based Discrete Firefly Algorithm for Constraint Satisfaction Problems

Authors:

Mahdi Bidar, Malek Mouhoub and Samira Sadaoui

Abstract: Constraint Satisfaction Problems (CSPs) are known NP-complete problems requiring systematic search methods of exponential time costs for solving them. To overcome this limitation, an alternative is to use metaheuristics. However, these techniques often suffer from immature convergence, and this is mainly due to a lack of adequate diversity of the potential solutions. To address this challenge, we update the Discrete Firefly Algorithm (DFA) with the Chaos Theory. We call the Chaotic Discrete Firefly Algorithm (CDFA) this proposed algorithm. To assess the performance in practice of the proposed CDFA, we conducted several experiments on CSP instances randomly generated based on the model RB. The results of the experiments demonstrate the efficiency of CDFA in dealing with CSPs.

Paper Nr: 18
Title:

On Enumerating All the Minimal Models for Particular CNF Formula Classes

Authors:

Yakoub Salhi

Abstract: In this work, we propose approaches for enumerating all the minimal models for two particular classes of CNF formulæ. The first class is that of PN formulæ which are defined as CNF formulæ where each clause is either positive or negative, whereas the second class is that of PH formulæ in which each clause is either positive or a Horn clause. We first provide an approach for enumerating all the minimal models in the case of PN formulæ that is based on the use of an algorithm for generating the minimal transversals of a hypergraph. We also propose a SAT-based encoding for solving the same problem. Then, we provide a characterization of the minimal models in the case of PH formulæ, which allows us to use our approaches in the case of PN formulæ for solving the problem of minimal model enumeration for PH formulæ. Finally, we describe an application in datamining of the problem of enumerating the minimal models in the case of PN formulæ.
Download

Paper Nr: 19
Title:

Implementation of Trajectory Planning for Automated Driving Systems using Constraint Logic Programming

Authors:

Christian Wriedt and Christoph Beierle

Abstract: Automated driving systems are a maturing technology that is considered to have a significant impact on mobility. Trajectory Planning is a safety-critical task that plays an important role in automated driving systems. In this paper, we present the implementation of a trajectory planning module called CLPTP (CLPTRAJECTORYPLANNER) using constraint logic programming (CLP) and evaluate it in simulated traffic situations. CLP allows us to express the constraints of the problem of trajectory planning in a declarative way. The approach makes the code less complex and more readable for domain experts compared to code using an imperative programming language. Compared to approaches making use of neural networks to manage the complexity of the problem of trajectory planning, the results of CLPTP are more comprehensible and easier to verify. Thus CLPTP can be seen as a step towards solving the problem of trajectory planning with explainable artificial intelligence. An evaluation of the execution time performance of our implementation shows that further research is needed to apply the approach in real world vehicles.
Download

Paper Nr: 21
Title:

Transforming the Emotion in Speech using a Generative Adversarial Network

Authors:

Kenji Yasuda, Ryohei Orihara, Yuichi Sei, Yasuyuki Tahara and Akihiko Ohsuga

Abstract: In recent years, natural and highly accurate outputs in domain transfer tasks have been achieved by deep learning techniques. Especially, the advent of Generative Adversarial Networks (GANs) has enabled the transfer of objects between unspecified domains. Voice conversion is a popular example of speech domain transfer, which can be paraphrased as domain transfer of speakers. However, most of the voice conversion studies have focused only on transforming the identities of speakers. Understanding other nuances in the voice is necessary for natural speech synthesis. To resolve this issue, we transform the emotions in speech by the most promising GAN model, CycleGAN. In particular, we investigate the usefulness of speech with low emotional intensity as training data. Such speeches are found to be useful when the training data contained multiple speakers.
Download

Paper Nr: 25
Title:

Acceleration of Backpropagation Training with Selective Momentum Term

Authors:

Diego M. Carvalho and Areolino A. Neto

Abstract: In many cases it is very hard to get an Artificial Neural Network (ANN) suitable for learning the solution, i.e., it cannot acquire the desired knowledge or needs an enormous number of training iterations. In order to improve the learning of ANN type Multi-Layer Perceptron (MLP), this work describes a new methodology for selecting weights, which will have the momentum term added to variation calculus of their values during each training iteration via Backpropagation (BP) algorithm. For that, the Pearson or Spearman correlation coefficients are used. Even very popular, the usage of BP algorithm has some drawbacks, among them the high convergence time is highlighted. A well-known technique used to reduce this disadvantage is the momentum term, which tries to accelerate the ANN learning keeping its stability, but when it is applied in all weights, as commonly used, with inadequate parameters, the result can be easily a failure in the training or at least an insignificant reduction of the ANN training time. The use of the Selective Momentum Term (SMT) can reduce the training time and, therefore, be also used for improving the training of deep neural networks.
Download

Paper Nr: 28
Title:

Can Machine Learning Predict Soccer Match Results?

Authors:

Giovanni Capobianco, Umberto Di Giacomo, Francesco Mercaldo, Vittoria Nardone and Antonella Santone

Abstract: Sport result prediction proposes an interesting challenge considering as popular and widespread are sport games, for instance tennis and soccer. The outcome prediction is a difficult task because there are a lot of factors that can afflict the final results and most of them are related to the player human behaviour. In this paper we propose a new feature set (related to the match and to players) aimed to model a soccer match. The set is related to characteristics obtainable not only at the end of the match, but also when the match is in progress. We consider machine learning techniques to predict the results of the match and the number of goals, evaluating a dataset of real-world data obtained from the Italian Serie A league in the 2017-2018 season. Using the RandomForest algorithm we obtain a precision of 0.857 and a recall of 0.750 in won match prediction, while for the goal prediction we obtain a precision of 0.879 in the number of goal prediction less than two, and a precision of 0.8 in the number of goal prediction equal or greater to two.
Download

Paper Nr: 29
Title:

Visibility Forecast for Airport Operations by LSTM Neural Network

Authors:

Tuo Deng, Aijie Cheng, Wei Han and Hai-Xiang Lin

Abstract: Visibility forecast is a meteorological problems which has direct impact to daily lives. For instance, timely prediction of low visibility situations is very important for the safe operation in airports and highways. In this paper, we investigate the use of Long Short-Term Memory(LSTM) model to predict visibility. By adjusting the loss function and network structure, we optimize the original LSTM model to make it more suitable for practical applications, which is superior to previous models in short-term low visibility prediction. In addition, there is a ”time delay problem” when the number of hours time ahead we try to forecast becomes larger, this problem is persistent given the limited amount of available training data. We report our attempt of applying re-sampling to deal with the time delay problem, and we find that this method can improve the accuracy of visibility prediction, especially for the low visibility case.
Download

Paper Nr: 37
Title:

Uncertain Formal Concept Analysis for the Analyze of a Course Satisfaction Questionnaire

Authors:

Guillaume Petiot

Abstract: The Formal Concept Analysis (FCA) is a method of data analysis often used in data mining. This method proposes to build a collection of formal concepts from a set of objects and their properties. These formal concepts can be ordered to provide a concept lattice. Several researches have demonstrated a link between the possibility theory and the formal concept analysis. Thus, it is possible to take into account the uncertainties of the properties by using the possibility theory before propagating them during the computation of the formal concepts. We propose in this paper an experimentation of the uncertain formal concept analysis for the extraction of knowledge in a satisfaction questionnaire for a course of professionalization in bachelor. Some questions can be open questions where the answers provided by students are given freely. For this purpose, we perform a text mining processing in order to categorize and classify the answers. During this processing, uncertainties can appear. In this research, we will handle these uncertainties by using the uncertain formal concept analysis. Then, we will extract the uncertain formal concepts from the concept lattice by using queries and represent the reduced lattice concepts with a visualization tool.
Download

Paper Nr: 43
Title:

Analogy-based Matching Model for Domain-specific Information Retrieval

Authors:

Myriam Bounhas and Bilel Elayeb

Abstract: This paper describes a new matching model based on analogical proportions useful for domain-specific Information Retrieval (IR). We first formalize the relationship between documents terms and query terms through analogical proportions and we propose a new analogical inference to evaluate document relevance for a given query. Then we define the analogical relevance of a document in the collection by aggregating two scores: the Agreement, measured by the number of common terms, and the Disagreement, measured by the number of different terms. The disagreement degree is useful to filter documents out from the response (retrieved documents), while the agreement score is convenient for document relevance confirmation. Experiments carried out on three IR Glasgow test collections highlight the effectiveness of the model if compared to the known efficient Okapi IR model.
Download

Paper Nr: 44
Title:

A Comparative Assessment of Ontology Weighting Methods in Semantic Similarity Search

Authors:

Antonio De Nicola, Anna Formica, Michele Missikoff, Elaheh Pourabbas and Francesco Taglino

Abstract: Semantic search is the new frontier for the search engines of the last generation. Advanced semantic search methods are exploring the use of weighted ontologies, i.e., domain ontologies where concepts are associated with weights, inversely related to their selective power. In this paper, we present and assess four different ontology weighting methods, organized according to two groups: intensional methods, based on the sole ontology structure, and extensional methods, where also the content of the search space is considered. The comparative assessment is carried out by embedding the different methods within the semantic search engine SemSim, based on weighted ontologies, and then by running four retrieval tests over a search space we have previously proposed in the literature. In order to reach a broad audience of readers, the key concepts of this paper have been presented by using a simple taxonomy, and the already experimented dataset.
Download

Paper Nr: 46
Title:

Using Proof Failures to Help Debugging MAS

Authors:

Bruno Mermet and Gaële Simon

Abstract: For several years, we have worked on the usage of theorem proving techniques to validate Multi-Agent Systems. In this article, we present a preliminary case study, that is part of larger work whose long-term goal is to determine how proof tools can be used to help to develop error-free Multi-Agent Systems. This article describes how an error caused by a synchronisation problem between several agents can be identified by a proof failure. We also show that analysing proof failures can help to find bugs that may occur only in a very particular context, which makes it difficult to analyse by standard debugging techniques.
Download

Paper Nr: 47
Title:

Efficient SAT Encodings for Hierarchical Planning

Authors:

Dominik Schreiber, Damien Pellier, Humbert Fiorino and Tomáš Balyo

Abstract: Hierarchical Task Networks (HTN) are one of the most expressive representations for automated planning problems. On the other hand, in recent years, the performance of SAT solvers has been drastically improved. To take advantage of these advances, we investigate how to encode HTN problems as SAT problems. In this paper, we propose two new encodings: GCT (Grammar-Constrained Tasks) and SMS (Stack Machine Simulation), which, contrary to previous encodings, address recursive task relationships in HTN problems. We evaluate both encodings on benchmark domains from the International Planning Competition (IPC), setting a new baseline in SAT planning on modern HTN domains.
Download

Paper Nr: 68
Title:

Data-driven Identification of Causal Dependencies in Cyber-Physical Production Systems

Authors:

Kaja Balzereit, Alexander Maier, Björn Barig, Tino Hutschenreuther and Oliver Niggemann

Abstract: Cyber-Physical Systems (CPS) are systems that connect physical components with software components. CPS used for production are called Cyber-Physical Production Systems (CPPS). Since the complexity of these systems can be very high, finding the cause of an error takes a lot of effort. In this paper, a data-driven approach to identify causal dependencies in cyber-physical production systems (CPPS) is presented. The approach is based on two different layers of learning algorithms: one low-level layer that processes the direct machine data and a higher-level learning layer that processes the output of the low-level layer. The low-level layer is based on different learning modules that can process differently typed data (continuous, discrete or both). The high-level learning algorithms are based on rule-based and case-based reasoning. Thus, causal dependencies are detected allowing the plant operator to find the error cause quickly.
Download

Paper Nr: 73
Title:

A Belief Approach for Detecting Spammed Links in Social Networks

Authors:

Salma Ben Dhaou, Mouloud Kharoune, Arnaud Martin and Boutheina Ben Yaghlane

Abstract: Nowadays, we are interconnected with people whether professionally or personally using different social networks. However, we sometimes receive messages or advertisements that are not correlated to the nature of the relation established between the persons. Therefore, it became important to be able to sort out our relationships. Thus, based on the type of links that connect us, we can decide if this last is spammed and should be deleted. Thereby, we propose in this paper a belief approach in order to detect the spammed links. Our method consists on modelling the belief that a link is perceived as spammed by taking into account the prior information of the nodes, the links and the messages that pass through them. To evaluate our method, we first add some noise to the messages, then to both links and messages in order to distinguish the spammed links in the network. Second, we select randomly spammed links of the network and observe if our model is able to detect them. The results of the proposed approach are compared with those of the baseline and to the k-nn algorithm. The experiments indicate the efficiency of the proposed model.
Download

Paper Nr: 87
Title:

New Indicator for Centrality Measurements in Passing-network Analysis of Soccer

Authors:

Masatoshi Kanbata, Ryohei Orihara, Yuichi Sei, Yasuyuki Tahara and Akihiko Ohsuga

Abstract: A number of fields including business, science, and sports, make use of data analytics. The evaluation of players and teams affect how tactics, training, and scouting are conducted in soccer teams. Data such as the number of shots and goals in match results are often used to evaluate players and teams. However, this is not enough to fully understand the potential of the players and teams. In this paper, we describe a new analysis method using passing-distribution data from soccer games. To evaluate the performance of players and teams, we applied graph mining. We also used an index called centrality, which evaluates individual contributions with an organization. In this research, we propose a new centrality model to improve existing conventional models. In the calculating the centrality of a given player pair, we consider not only the shortest sequence of passing but also longer ones. In this research, we verified the significance of these indicators by applying the data of UEFA EURO 2008, 2012, and 2016. As a result, we found our method to be more consistent with game results than conventional methods.
Download

Paper Nr: 88
Title:

On Mining Conditions using Encoder-decoder Networks

Authors:

Fernando O. Gallego and Rafael Corchuelo

Abstract: A condition is a constraint that determines when something holds. Mining them is paramount to understanding many sentences properly. There are a few pattern-based approaches that fall short because the patterns must be handcrafted and it is not easy to characterise unusual ways to express conditions; there is one machine-learning approach that requires specific-purpose dictionaries, taxonomies, and heuristics, works on opinion sentences only, and was evaluated on a small dataset with Japanese sentences on hotels. In this paper, we present an encoder-decoder model to mine conditions that does not have any of the previous drawbacks and outperforms the state of the art in terms of effectiveness.
Download

Paper Nr: 90
Title:

System Implementation for the Detection of Weak Signals of the Future in Heterogeneous Documents by Text Mining and Natural Language Processing Techniques

Authors:

Israel Griol-Barres, Sergio Milla and José Millet

Abstract: Not being able to cope with the constant changes in the market is currently one of the biggest threats for companies and start-ups. Therefore, the development of new systems to detect significant phenomena and future changes, is a key component for correct decision making that sets a correct course in the organisation. For this reason, a business intelligence architecture system is hereby proposed to allow the detection of discrete changes or weak signals in the present, indicative of more significant phenomena and transcendental changes in the future. In contrast to work currently available focusing on structured information sources, or at most with a single type of data source, the detection of these signals is here quantitatively based on heterogeneous and unstructured documents of various kinds (scientific journals, newspaper articles and social networks), to which text mining and natural language processing techniques (a multi-word expression analysis) are applied. The system has been tested to study the future of solar panels and the artificial intelligence sectors, obtaining promising results to help business experts in the recognition of new driving factors of their markets and the development of new opportunities.
Download

Paper Nr: 92
Title:

Placement-and-Profit-Aware Association Rules Mining

Authors:

Runyu Ma, Hantian Li, Jin Cen and Amrinder Arora

Abstract: Previous approaches on association rule mining in recommendation have already achieved promising performances. However, to the best of our knowledge, they seldom simultaneously take the profit and placement factor into consideration. In E-commerce recommendation scenario, the order of the recommendation reflects as placement. In this paper, we propose a novel placement-and-profit-aware association rule mining algorithm to maximize profit as well as maintaining recommendation accuracy. We also propose two metrics: Expectation of Profit (EOP), which measures the overall profit, and Expectation of Click rate (EOC), which measures the user experience. Experiments on SPMF dataset show that the proposed algorithm can improve the EOP significantly with only slight decrease in EOC.
Download

Paper Nr: 100
Title:

Breast Cancer Detection using Deep Convolutional Neural Network

Authors:

Hana Mechria, Mohamed S. Gouider and Khaled Hassine

Abstract: Deep Convolutional Neural Network (DCNN) is considered as a popular and powerful deep learning algorithm in image classification. However, there are not many DCNN applications used in medical imaging, because large dataset for medical images is not always available. In this paper, we present two DCNN architectures, a shallow DCNN and a pre-trained DCNN model: AlexNet, to detect breast cancer from 8000 mammographic images extracted from the Digital Database for Screening Mammography. In order to validate the performance of DCNN in breast cancer detection using a big data , we carried out a comparative study with a second deep learning algorithm Stacked AutoEncoders (SAE) in terms accuracy, sensitivity and specificity. The DCNN method achieved the best results with 89.23% of accuracy, 91.11% of sensitivity and 87.75% of specificity.
Download

Paper Nr: 102
Title:

Probability based Proof Number Search

Authors:

Zhang Song, Hiroyuki Iida and H. J. van den Herik

Abstract: Probability based proof number search (PPN-search) is a game tree search algorithm improved from proof number search (PN-search) (Allis et al., 1994), with applications in solving games or endgame positions. PPN-search uses one indicator named “probability based proof number” (PPN) to indicate the “probability” of proving a node. The PPN of a leaf node is derived from Monte-Carlo evaluations. The PPN of an internal node is backpropagated from its children following AND/OR probability rules. For each iteration, PPN-search selects the child with the maximum PPN at OR nodes and minimum PPN at AND nodes. This holds from the root to a leaf. The resultant node is considered to be the most proving node for expansion. In this paper, we investigate the performance of PPN-search on P-game trees (Kocsis and Szepesvári, 2006) and compare our results with those from other game solvers such as MCPN-search (Saito et al., 2006), PN-search, the UCT solver (Winands et al., 2008), and the pure MCTS solver (Winands et al., 2008). The experimental results show that (1) PPN-search takes less time and fewer iterations to solve a P-game tree on average, and (2) the error rate of selecting a correct solution decreases faster and more smoothly as the iteration number increases.
Download

Paper Nr: 103
Title:

Strategies for Runtime Prediction and Mathematical Solvers Tuning

Authors:

Michael Barry and René Schumann

Abstract: Mathematical solvers have evolved to become complex software and thereby have become a difficult subject for Runtime Prediction and parameter tuning. This paper studies various Machine Learning methods and data generation techniques to compare their effectiveness for both Runtime Prediction and parameter tuning. We show that machine Learning methods and Data Generation strategies that perform well for Runtime Prediction do not necessary result in better results for solver tuning. We show that Data Generation algorithms with an emphasis on exploitation combined with Random Forest is successful and random trees are effective for Runtime Prediction. We apply these methods to a hydro power model and present results from two experiments.
Download

Paper Nr: 111
Title:

CollabNet - Collaborative Deep Learning Network

Authors:

Moisés F. Lima Junior, Will M. Almeida and Areolino A. Neto

Abstract: The goal is an improvement on learning of deep neural networks. This improvement is here called the CollabNet network, which consists of a new method of insertion of new layers hidden in deep feedforward neural networks, changing the traditional way of stacking autoencoders. The new form of insertion is considered collaborative and seeks to improve the training against approaches based on stacked autoencoders. In this new approach, the addition of a new layer is carried out in a coordinated and gradual way, keeping under the control of the designer the influence of this new layer in training and no longer in a random and stochastic way as in the traditional stacking. The collaboration proposed in this work consists of making the learning of newly inserted layer continuing the learning obtained from previous layers, without prejudice to the global learning of the network. In this way, the freshly added layer collaborates with the previous layers and the set works in a way more aligned to the learning. CollabNet has been tested in the Wisconsin Breast Cancer Dataset database, obtaining a satisfactory and promising result.
Download

Paper Nr: 112
Title:

Towards Adaptive Deep Reinforcement Game Balancing

Authors:

Ashey Noblega, Aline Paes and Esteban Clua

Abstract: The experience of a player regarding the difficulty of a video game is one of the main reasons for he/she decide to keep playing the game or abandon it. Effectively, player retention is one of the primary concerns related to the game development process. However, the experience of a player with a game is unique, making impractical to anticipate how they will face the gameplay. This work leverages the recent advances in Reinforcement Learning (RL) and Deep Learning (DL) to create intelligent agents that are able to adapt to the abilities of distinct players. We focus on balancing the difficulty of the game based on the information that the agent observes from the 3D environment as well as the current state of the game. In order to design an agent that learns how to act while still maintaining the balancing, we propose a reward function based on a balancing constant. We require that the agent remains inside a range around this constant during the training. Our experimental results show that by using such a reward function and combining information from different types of players it is possible to have adaptable agents that fit the player.
Download

Paper Nr: 116
Title:

Question and Answer Classification in Czech Question Answering Benchmark Dataset

Authors:

Daša Kušniráková, Marek Medved and Aleš Horák

Abstract: In this paper, we introduce a new updated version of the Czech Question Answering database SQAD v2.1 (Simple Question Answering Database) with the update being devoted to improved question and answer classification. The SQAD v2.1 database contains more than 8,500 question-answer pairs with all appropriate metadata for QA training and evaluation. We present the details and changes in the database structure as well as a new algorithm for detecting the question type and the actual answer type from the text of the question. The algorithm is evaluated with more than 4,000 question answer pairs reaching the F1-measure of 88% for question typed and 85% for answer type detection.
Download

Paper Nr: 121
Title:

Dealing with Permanent Agent Failure in Dynamic Agents Organisations

Authors:

Asia S. Al-Karkhi and Maria Fasli

Abstract: This paper is an exploratory study that focuses on the creation of open and dynamic agent organisations which can sustain the provision of services to requesting customers in the presence of agent failure. In a distributed environment, agents within networks and organisations are prone to failure. This can inevitably lead to decreases in the individual agents’ utilisation as well as in the whole system’s and to the loss of tasks. Here, we present an approach to tackling this challenge by enabling agents to prevent these kinds of disruptions. In an environment where agents create organisations to increase the execution of tasks, we employ the Henchman Recovery Protocol (HRP) within each organisation; this enables the agents within an organisation to maintain its functionality in the presence of agent failures and in particular in the case of the lead (Head) of the organisation failing. Furthermore, we explore the stability and evolution of organisations over a period of time and when agents drop out of organisations (due to permanent failure) while new agents may enter the environment and either join existing organisations or create new ones. We conduct our study in the context of a grid-like computing system which was implemented in the Repast Simphony agent-based simulation environment.
Download

Paper Nr: 135
Title:

Normalizing Emotion-Driven Acronyms towards Decoding Spontaneous Short Text Messages

Authors:

Bizhanova Aizhan and Atsushi Fujii

Abstract: Reflecting the rapid growth in the use of Social Networking Services (SNSs), it has of late become popular for users to share their feelings, impression, and opinions with each other, about what they saw or experienced, rapidly by means of short text messages (SMS). This trend has let a large number of users consciously or unconsciously use emotion-bearing words and also acronyms to reduce the number of characters to type. We have noticed this new emerging category of language unit, namely “Emotion-Driven Acronyms (EDAs)”. Because by definition, each acronym consists of less characters than its original full form, the acronyms for different full forms often coincidently identical. Consequently, the misuse of EDAs substantially decreases the readability of messages. Our long-term research goal is to normalize text in a corrupt language into the canonical one. In this paper, as the first step towards the exploration of EDAs, we focus only on the normalization for EDAs and propose a method to disambiguate the occurrence of an EDA that corresponds to different full forms depending on the context, such as “smh (so much hate / shaking my head)”. We also demonstrate what kind of features are effective in our task experimentally and discuss the nature of EDAs from different perspectives.
Download

Paper Nr: 139
Title:

Type-Theory of Acyclic Algorithms with Generalised Immediate Terms

Authors:

Roussanka Loukanova

Abstract: The paper extends the higher-order, type-theory Lλ ar of acyclic algorithms by classifying some explicit terms as generalised immediate terms. We introduce a restricted, specialised iβ-rule and reduction of generalised immediate terms to their iβ-canonical forms. Then, we incorporate the iβ-rule in the reduction calculus of Lλ ar. The new reduction calculus provides more efficient algoritms for computation of values of terms, by using iterative calculations.
Download

Paper Nr: 143
Title:

Decomposing Constraint Satisfaction Problems by Means of Meta Constraint Satisfaction Optimization Problems

Authors:

Sven Löffler, Ke Liu and Petra Hofstedt

Abstract: This paper describes a new approach to decompose constraint satisfaction problems (CSPs) using an auxiliary constraint satisfaction optimization problem (CSOP) that detects sub-CSPs which share only few common variables. The purpose of this approach is to find sub-CSPs which can be solved in parallel and combined to a complete solution of the original CSP. Therefore, our decomposition approach has two goals: 1. to evenly balance the workload distribution over all cores and solve the partial CSPs as fast as possible and 2. to minimize the number of shared variables to make the join process of the solutions as fast as possible.
Download

Paper Nr: 149
Title:

Integrating an Association Rule Mining Agent in an ERP System: A Proposal and a Computational Scalability Analysis

Authors:

Rafael Marin Machado de Souza, Fabrício G. Vilasbôas, Pollyana Notargiacomo and Leandro Nunes de Castro

Abstract: Deployment flexibility, low development cost, and value-adding tools are some of the features that developers are looking for in ERP systems. Modularization through software agents is one way of achieving these objectives. In this sense, the present paper proposes the planning, implementation and integration of a software agent for association rule mining into an ERP system. The development and use of tools for all Knowledge Discovery in Databases (KDD) phases (pre-processing, data mining and post-processing), will be presented. This includes input data, file loading for the agent processing, use of the Apriori association rule mining algorithm, generation of output files with association rules, use of agent outputs for database storage and use of the stored data by the item recommendation tool. Experiments were carried out focusing the assessment of the running profile for databases of different sizes and using different computational architectures.
Download

Paper Nr: 158
Title:

Air Quality Forecast through Integrated Data Assimilation and Machine Learning

Authors:

Hai X. Lin, Jianbing Jin and Jaap van den Herik

Abstract: Numerical models of chemical transport have been used to simulate the complex processes involved in the formation and transport of air pollutants. Although these models can predict the spatiotemporal variability of a variety of chemical species, the accuracy of these models is often limited. Therefore, in the past two decades, data assimilation methods have been applied to use the available measurements for improving the forecast. Nowadays, machine learning techniques provide new opportunities for improving the air quality forecast. A case study on PM10 concentrations during a dust storm is performed. It is known that the PM10 concentrations are caused by multiple emission sources, e.g., dust from desert and anthropogenic emissions. An accurate modeling of the PM10 concentration levels owing to the local anthropogenic emissions is essential for an adequate evaluation of the dust level. However, real-time measurement of local emissions is not possible, so no direct data is available. Actually, the lack of in-time emission inventories is one of the main reasons that current numerical chemical transport models cannot produce accurate anthropogenic PM10 simulations. Using machine learning techniques to generate local emissions based on real-time observations is a promising approach. We report how it can be combined with data assimilation to improve the accuracy of air quality forecast considerably.
Download

Paper Nr: 160
Title:

Fake News Detection via NLP is Vulnerable to Adversarial Attacks

Authors:

Zhixuan Zhou, Huankang Guan, Meghana M. Bhat and Justin Hsu

Abstract: News plays a significant role in shaping people’s beliefs and opinions. Fake news has always been a problem, which wasn’t exposed to the mass public until the past election cycle for the 45th President of the United States. While quite a few detection methods have been proposed to combat fake news since 2015, they focus mainly on linguistic aspects of an article without any fact checking. In this paper, we argue that these models have the potential to misclassify fact-tampering fake news as well as under-written real news. Through experiments on Fakebox, a state-of-the-art fake news detector, we show that fact tampering attacks can be effective. To address these weaknesses, we argue that fact checking should be adopted in conjunction with linguistic characteristics analysis, so as to truly separate fake news from real news. A crowdsourced knowledge graph is proposed as a straw man solution to collecting timely facts about news events.
Download

Paper Nr: 163
Title:

Towards Minimizing e-Commerce Returns for Clothing

Authors:

A. K. Seewald, T. Wernbacher, A. Pfeiffer, N. Denk, M. Platzer, M. Berger and T. Winter

Abstract: The importance of e-commerce including the associated freight traffic with all its negative consequences (e.g. congestion, noise, emissions) is constantly increasing. Already in 2015, an European market volume of 444 billion Euros at an annual growth of 13.3% was achieved, of which clothing and footwear account for 12.7% as the largest category (Willemsen et al., 2016). However, online commerce will only have a better footprint than buying in the local retail shop under optimal conditions (for example: group orders, always present at home delivery, no returns and no same day delivery). Next to frequent single deliveries, CO2 intensive and underutilized transport systems, returned goods are the main problem of online shopping. The last is currently estimated at up to 50% (Hofacker and Langenberg, 2015; Kristensen et al., 2013). Our research project Think!First tackles these problems in freight mobility by using an unique combination of gamification elements, persuasive design principles and machine learning. Customers are animated, targeted and nudged to choose effective and sustainable means of transport when shopping online while ensuring best fit by compensating both manufacturer and customer biases in body size estimation. Here we show preliminary results and also present a slightly modified rule learning algorithm that always characterizes a given class (here: returns).
Download

Paper Nr: 164
Title:

Towards Controlled Transformation of Sentiment in Sentences

Authors:

Wouter Leeftink and Gerasimos Spanakis

Abstract: An obstacle to the development of many natural language processing products is the vast amount of training examples necessary to get satisfactory results. The generation of these examples is often a tedious and time-consuming task. This paper this paper proposes a method to transform the sentiment of sentences in order to limit the work necessary to generate more training data. This means that one sentence can be transformed to an opposite sentiment sentence and should reduce by half the work required in the generation of text. The proposed pipeline consists of a sentiment classifier with an attention mechanism to highlight the short phrases that determine the sentiment of a sentence. Then, these phrases are changed to phrases of the opposite sentiment using a baseline model and an autoencoder approach. Experiments are run on both the separate parts of the pipeline as well as on the end-to-end model. The sentiment classifier is tested on its accuracy and is found to perform adequately. The autoencoder is tested on how well it is able to change the sentiment of an encoded phrase and it was found that such a task is possible. We use human evaluation to judge the performance of the full (end-to-end) pipeline and that reveals that a model using word vectors outperforms the encoder model. Numerical evaluation shows that a success rate of 54.7% is achieved on the sentiment change.
Download

Paper Nr: 167
Title:

On-the-spot Knowledge Refinement for an Interactive Recommender System

Authors:

Yuichiro Ikemoto and Kazuhiro Kuwabara

Abstract: This paper proposes a method to refine knowledge about items in an item database for an interactive recommender system. The proposed method is integrated into a recommender system and invoked when the system recognizes a problem with the item database from users’ feedback about recommended items. The proposed method collects information from a user via similar interactions to those of a recommendation process. In this way, a user who is knowledgeable in a target domain, but does not necessarily know the internal system can participate in the knowledge refinement process. Thus, the proposed method paves the way for applying crowdsourcing to knowledge refinement.
Download

Paper Nr: 170
Title:

Interactive Lungs Auscultation with Reinforcement Learning Agent

Authors:

Tomasz Grzywalski, Riccardo Belluzzo, Szymon Drgas, Agnieszka Cwalińska and Honorata Hafke-Dys

Abstract: To perform a precise auscultation for the purposes of examination of respiratory system normally requires the presence of an experienced doctor. With most recent advances in machine learning and artificial intelligence, automatic detection of pathological breath phenomena in sounds recorded with stethoscope becomes a reality. But to perform a full auscultation in home environment by layman is another matter, especially if the patient is a child. In this paper we propose a unique application of Reinforcement Learning for training an agent that interactively guides the end user throughout the auscultation procedure. We show that intelligent selection of auscultation points by the agent reduces time of the examination fourfold without significant decrease in diagnosis accuracy compared to exhaustive auscultation.
Download

Paper Nr: 173
Title:

End-to-end Learning Approach for Autonomous Driving: A Convolutional Neural Network Model

Authors:

Yaqin Wang, Dongfang Liu, Hyewon Jeon, Zhiwei Chu and Eric T. Matson

Abstract: End-to-end approach is one of the frequently used approaches for the autonomous driving system. In this study, we adopt the end-to-end approach because this approach has been approved to lead to a distinguished performance with a simpler system. We build a convolutional neural network (CNN) to map raw pixels from cameras of three different angles and to generate steering commands to drive a car in the Udacity simulator. Our proposed model has a promising result, which is more accurate and has lower loss rate comparing to previous models.
Download

Paper Nr: 174
Title:

Probabilistic Method for Estimation of Spinning Reserves in Multi-connected Power Systems with Bayesian Network-based Rescheduling Algorithm

Authors:

Yerzhigit Bapin and Vasileios Zarikas

Abstract: This study proposes a new stochastic spinning reserve estimation model applicable to multi-connected energy systems with reserve rescheduling algorithm based on Bayesian Networks. The general structure of the model is developed based on the probabilistic reserve estimation model that considers random generator outages as well as load and renewable energy forecast errors. The novelty of the present work concerns the additional Bayesian layer which is linked to the general model. It conducts reserve rescheduling based on the actual net demand realization and other reserve requirements. The results show that the proposed model improves estimation of reserve requirements by reducing the total cost of the system associated with reserve schedule.
Download

Paper Nr: 180
Title:

Designing Transparent and Autonomous Intelligent Vision Systems

Authors:

Joanna I. Olszewska

Abstract: To process vast amounts of visual data such as images, videos, etc. in an automatic and computationally efficient way, intelligent vision systems have been developed over the last three decades. However, with the increasing development of complex technologies like companion robots which require advanced machine vision capabilities and, on the other hand, the growing attention to data security and privacy, the design of intelligent vision systems faces new challenges such as autonomy and transparency. Hence, in this paper, we propose to define the main requirements for the new generation of intelligent vision systems (IVS) we demonstrated in a prototype.
Download

Paper Nr: 184
Title:

An Alternative to Restricted-Boltzmann Learning for Binary Latent Variables based on the Criterion of Maximal Mutual Information

Authors:

David Edelman

Abstract: The latent binary variable training problem used in the pre-training process for Deep Neural Networks is approached using the Principle (and related Criterion) of Maximum Mutual Information (MMI). This is presented as an alternative to the most widely-accepted ’Restricted Boltzmann Machine’ (RBM) approach of Hinton. The primary contribution of the present article is to present the MMI approach as the arguably more logically ’natural’ and logically simple means to the same ends. Additionally, the relative ease and effectiveness of the approach for application will be demonstrated for an example case.
Download

Paper Nr: 187
Title:

Towards Locative Inconsistency-tolerant Hierarchical Probabilistic CTL Model Checking: Survey and Future Work

Authors:

Norihiro Kamide and Juan A. Bernal

Abstract: A locative inconsistency-tolerant hierarchical probabilistic computation tree logic (LIHpCTL) is introduced in this paper to establish the logical foundation of a new model checking paradigm. This logic is an extension of several previously proposed extensions of the standard temporal logic known as CTL, which is widely used for model checking. The extended model checking paradigm proposed is intended to appropriately verify locative (spatial), inconsistent, hierarchical, probabilistic (randomized), and time-dependent concurrent systems. Additionally, a survey of various studies on probabilistic, inconsistency-tolerant, and hierarchical temporal logics and their applications in model checking is conducted.
Download

Paper Nr: 188
Title:

Deep Separable Convolution Neural Network for Illumination Estimation

Authors:

Minquan Wang and Zhaowei Shang

Abstract: Illumination estimation has been studied for a long time. The algorithms to solve the problem can be roughly divided into two categories: statistical-based and learning-based. Statistical-based algorithm has the advantage of fast computing speed but low accuracy. Learning-based algorithm improve the estimation accuracy to some extent, but generally have high computational complexity and storage space. In this paper, a new deep convolution neural network is proposed. We design the network with more layers (11 convolution layers) than the existing methods, remove the “skip connection” and “Global Average Pooling” is used to replace “Fully Connection” layer which is commonly used in the existing methods. We use the separable convolution instead of the standard convolution to reduce the number of parameters. In reprocessed Color Checker Dataset, compared with the present state-of-the-art the proposed method reduces the average angular error by about 60%. At the same time, using separable convolution and “Global Average Pooling” reduces the number of parameters by about 86% compared with do not use them.
Download

Paper Nr: 190
Title:

Contextual Action with Multiple Policies Inverse Reinforcement Learning for Behavior Simulation

Authors:

Nahum Alvarez and Itsuki Noda

Abstract: Machine learning is a discipline with many simulator-driven applications oriented to learn behavior. However, behavior simulation it comes with a number of associated difficulties, like the lack of a clear reward function, actions that depend of the state of the actor and the alternation of different policies. We present a method for behavior learning called Contextual Action Multiple Policy Inverse Reinforcement Learning (CAMP-IRL) that tackles those factors. Our method allows to extract multiple reward functions and generates different behavior profiles from them. We applied our method to a large scale crowd simulator using intelligent agents to imitate pedestrian behavior, making the virtual pedestrians able to switch between behaviors depending of the goal they have and navigating efficiently across unknown environments.
Download

Paper Nr: 192
Title:

Classification Model for Cerebral Aneurysm Rupture Prediction using Medical and Blood-flow-simulation Data

Authors:

Masaaki Suzuki, Toshiyuki Haruhara, Hiroyuki Takao, Takashi Suzuki, Soichiro Fujimura, Toshihiro Ishibashi, Makoto Yamamoto, Yuichi Murayama and Hayato Ohwada

Abstract: Stroke is a serious cerebrovascular condition, in which brain cells die due to an abrupt blockage of arteries supplying blood and oxygen or due to bleeding in the brain tissue when a blood vessel bursts or ruptures. Because stroke occurs suddenly in most people, prevention is oftentimes difficult. In Japan, this condition is one of the major causes of death, which is associated with high medical cost, especially among the society’s aging population. Therefore, stroke prediction and treatment is important. Stroke incidences can be avoided by a preventive treatment based on the risk of onset. However, since judgment of the onset risk largely depends on the individual experience and skill of the doctor, a highly accurate prediction method that is independent of the doctor’s experience and skill is the focus of this study. The target of prediction for this research is subarachnoid hemorrhage that is part of stroke. Logistic regression and support vector machine that predict cerebral aneurysm rupture by machine learning using combined medical data and cerebral blood-flow-simulation data were employed to analyze 338 cerebral aneurysm samples (35 ruptured, 303 unruptured). SMOTE algorithm solved the imbalance of data, while the SelectKBest algorithm was used to extract important features from the total 70 features obtained from both data. Out of the 27 important features extracted, 40% belonged to the medical data and the remaining 60% were from the blood-flow-simulation data. Using logistic regression as a classification model, we found the sensitivity of 0.64 and the specificity of 0.85. The results validated the possibility of a highly accurate method of cerebral aneurysm rupture prediction by machine learning using engineering information obtained from mechanical simulation.
Download

Paper Nr: 193
Title:

An Ontology-based Web Crawling Approach for the Retrieval of Materials in the Educational Domain

Authors:

Mohammed Ibrahim and Yanyan Yang

Abstract: As the web continues to be a huge source of information for various domains, the information available is rapidly increasing. Most of this information is stored in unstructured databases and therefore searching for relevant information becomes a complex task and the search for pertinent information within a specific domain is time-consuming and, in all probability, results in irrelevant information being retrieved. Crawling and downloading pages that are related to the user’s enquiries alone is a tedious activity. In particular, crawlers focus on converting unstructured data and sorting this into a structured database. In this paper, among others kind of crawling, we focus on those techniques that extract the content of a web page based on the relations of ontology concepts. Ontology is a promising technique by which to access and crawl only related data within specific web pages or a domain. The methodology proposed is a Web Crawler approach based on Ontology (WCO) which defines several relevance computation strategies with increased efficiency thereby reducing the number of extracted items in addition to the crawling time. It seeks to select and search out web pages in the education domain that matches the user’s requirements. In WCO, data is structured based on the hierarchical relationship, the concepts which are adapted in the ontology domain. The approach is flexible for application to crawler items for different domains by adapting user requirements in defining several relevance computation strategies with promising results.
Download

Paper Nr: 195
Title:

Multimodal Sentiment and Gender Classification for Video Logs

Authors:

Sadam Al-Azani and El-Sayed M. El-Alfy

Abstract: Sentiment analysis has recently attracted an immense attention from the social media research community. Until recently, the focus has been mainly on textual features before new directions are proposed for integration of other modalities. Moreover, combining gender classification with sentiment recognition is a more challenging problem and forms new business models for directed-decision making. This paper explores a sentiment and gender classification system for Arabic speakers using audio, textual and visual modalities. A video corpus is constructed and processed. Different features are extracted for each modality and then evaluated individually and in different combinations using two machine learning classifiers: support vector machines and logistic regression. Promising results are obtained with more than 90% accuracy achieved when using support vector machines with audio-visual or audio-text-visual features.
Download

Paper Nr: 12
Title:

Plug and Play Deep Convolutional Neural Networks

Authors:

Patrick Neary and Vicki Allan

Abstract: Major gains have been made in recent years in object recognition due to advances in deep convolutional neural networks. One struggle with deep learning is identifying an optimal network architecture for a given problem. Often different configurations are tried until one is identified that gives acceptable results. This paper proposes an asynchronous learning algorithm that finds an optimal network configuration by automatically adjusting network hyperparameters.
Download

Paper Nr: 20
Title:

A Causality Analysis for Nonlinear Classification Model with Self-Organizing Map and Locally Approximation to Linear Model

Authors:

Yasuhiro Kirihata, Takuya Maekawa and Takashi Onoyama

Abstract: In terms of nonlinear machine learning classifier such as Deep Learning, machine-learning model is generally a black box which has issue not to be clear the causality among its output classification and input attributes. In this paper, we propose a causality analysis method with self-organizing map and locally approximation to linear model. In this method, self-organizing map generates the cluster of input data and local linear models for each node on the map provides explanation of the generated model. Applying this method to the member rank prediction model based on Deep Learning, we validated our proposed method.
Download

Paper Nr: 24
Title:

A Meta Constraint Satisfaction Optimization Problem for the Optimization of Regular Constraint Satisfaction Problems

Authors:

Sven Löffler, Ke Liu and Petra Hofstedt

Abstract: This paper describes a new approach on optimization of regular constraint satisfaction problems (rCSPs) using an auxiliary constraint satisfaction optimization problem (CSOP) that detects areas with a potentially high number of conflicts. The purpose of this approach is to remove conflicts by the combination of regular constraints with intersection and concatenation of their underlying deterministic finite automatons (DFAs). This, eventually, often allows to significantly speed-up the solution process of the original rCSP.
Download

Paper Nr: 26
Title:

“Majorly Adapted Translator”: Towards Major Adaptation in ITS

Authors:

Tina Daaboul and Hicham Hage

Abstract: Culturally Aware Learning Systems are intelligent systems that adapt learning materials or techniques to the culture of learners having different “country, hobbies, experiences, etc.”, helping them better understand the topics being taught. In higher education, many learning sessions involve students of different majors. As observed, many instructors tend to manually modify the exercises several times, once for every major to adapt to the culture, which is tedious and impractical. Therefore, in this paper we propose an approach to making learning sessions adaptable to the major of the learner. Specifically, this work introduces an Artificial Intelligent system, “Majorly Adapted Translator (MAT)”, which aims at translating and adapting exercises from one major to another. MAT has two main phases, the first identifies the parts of an exercise that needs changing and creates an exercise template. The second translates and adapts the exercise. This work, highlights the first phase, the Feature Extract phase, which relies on our own relation extraction method to identify variables which extracts relations specific to named entities by using dependency relations and shallow parsing. Moreover, we report the performance of the system that was tested on a number of probability exercises.
Download

Paper Nr: 35
Title:

Comma Analysis and Processing for Improving Translation Quality of Long Sentences in Rule-based English-Korean Machine Translation

Authors:

Sung-Dong Kim

Abstract: Current English-Korean machine translation system cannot provide practical translation quality mainly due to the difficulties in long sentence parsing. Long sentences generally include commas, resulting in lots of different possible sentence structures. It is very difficult to accurately parse the long sentences that have commas. The roles of the commas in constructing sentences have to be identified and then the syntactic analysis should be performed according to the roles of the commas for accurate parsing of the long sentences. This paper presents the analysis results of the comma usages and the comma processing methods for each comma usage. And it also proposes the comma usage classification method using machine learning technique. In experiment, some improved translation results, by identifying comma usage and processing the commas, are also presented.
Download

Paper Nr: 36
Title:

An Investigation on the Data Mining to Develop Smart Tire

Authors:

Jae-Cheon Lee, Hao Liu, Young G. Seo, Seong W. Kwak, Ho S. Lee, Hae J. Jo and Sangsu Park

Abstract: A smart tire is required to improve driving safety for an intelligent vehicle especially for automated driving electric vehicles. It is necessary to provide information of tire contact forces (vertical, longitudinal, and lateral directions) to control velocity and steering angle of the autonomous vehicle so as to ensure driving stability. This study presents a smart tire system with the data mining to estimate the vertical load by using the tire deformation data in particular. Firstly, the hardware system construction of the smart tire in which tire deformation on driving by using strain gauge is described. And then the test condition is set up and total 27 sets of experimental data are processed to perform correlation analysis for specifications of measured waves. Next, the estimation algorithm of smart tire vertical load is derived by considering the area of tire-ground contact patch and also by introducing compensate coefficient of transverse direction length of contact area. The experimental results show the proposed estimation algorithm is feasible and precise. The advanced adaptive and precise estimation algorithm with artificial neural network will be developed further.
Download

Paper Nr: 45
Title:

HEART: Using Abstract Plans as a Guarantee of Downward Refinement in Decompositional Planning

Authors:

Antoine Gréa, Samir Aknine and Laetitia Matignon

Abstract: In recent years the ubiquity of artificial intelligence raised concerns among the uninitiated. The misunderstanding is further increased since most advances do not have explainable results. For automated planning, the research often targets speed, quality, or expressivity. Most existing solutions focus on one criteria while not addressing the others. However, human-related applications require a complex combination of all those criteria at different levels. We present a new method to compromise on these aspects while staying explainable. We aim to leave the range of potential applications as wide as possible but our main targets are human intent recognition and assistive robotics. We propose the HEART planner, a real-time decompositional planner based on a hierarchical version of Partial Order Causal Link (POCL). It cyclically explores the plan space while making sure that intermediary high level plans are valid and will return them as approximate solutions when interrupted. These plans are proven to be a guarantee of solvability. This paper aims to evaluate that process and its results compared to classical approaches in terms of efficiency and quality.
Download

Paper Nr: 48
Title:

Single Conspiracy Number Analysis in Checkers

Authors:

Sarot Busala, Zhang Song, Hiroyuki Iida, Mohd A. Khalid and Umi K. Yusof

Abstract: Game playing provides the medium for a variety of algorithms to formulate play decisions that surpass human expert. However, the reasons that distinguish the winning and losing positions remain actively researched which leads to the utilization of the search “indicator”. Conspiracy number search (CNS) and proof number search (PNS) had been popularly adopted as the search indicators in MIN/MAX and AND/OR tree, respectively. However, their limitations had encouraged the need for an alternative search indicator. The single conspiracy number (SCN) is a search indicator inspired by CNS and PNS which measure the difficulty of getting MIN/MAX value over a threshold point for a current root node. Recently, SCN had been successfully applied in Chinese chess to analyze both the progress pattern and long-term position. In this paper, analysis of the SCN within the game of checkers is conducted where different SCN threshold values and varying depth of the search tree were considered. Checkers was chosen due to its smaller search space and contain a rule that affects the shape of the search tree. The experimental results show that the SCN values stabilize as the depth of the search tree increases whether the player is in winning, drawing or losing position.
Download

Paper Nr: 52
Title:

A Cuckoo Search Algorithm for 2D-cutting Problem in Decorative Ceramic Production Lines with Defects

Authors:

Javier Monzon, Rony Cueva, Manuel Tupia and Mariuxi Bruzza

Abstract: The residues generated from ceramic cuttings are one of the major causes of waste in the ceramic production industry, with losses being around 40% of the used material. Hence, reducing residues from materials used is critical for lowering the production cost. It is worth mentioning that in this industry the material also shows high rates of defects, a constraint which most researches dealing with the 2D-cutting problem lack. This paper develops a bio-inspired metaheuristic called Cuckoo Search to solve the problem of exposed material cutting as an alternative solution to the genetic algorithm already developed by the authors, which will also be used to measure the Cuckoo Search algorithm performance.
Download

Paper Nr: 54
Title:

Prediction of Learning Improvement in Mathematics through a Video Game using Neurocomputational Models

Authors:

Richard Torres-Molina, Andrés Riofrío-Valdivieso, Carlos Bustamante-Orellana and Francisco Ortega-Zamorano

Abstract: Learning math is important for the academic life of students: the development of mathematical skills is influenced by different characteristics of students such as geographical position, economic level, parents’ education, achievement level, teacher objectives, social level, use of information and communication technologies by teachers, learner motivation, gender, age, preferences for playing video games, and the school year of the students. In this work, these previously mentioned characteristics were considered as the attributes (inputs) of a multilayer neural network that uses a backpropagation algorithm to predict the percentage of improvement in mathematics through a 2D mathematical video game that was developed to allow the children to practice addition and subtraction operations. After applying the neural model, using the twelve attributes mentioned before and the backpropagation algorithm, there was a network of one layer with ten neurons and another network of two layers with 5 neurons in the first layer and 20 neurons in the second layer. Both architectures produced a mean squared error smaller than 0.0069 in the prediction of the student’s percentage of improvement in mathematics, being the best configurations found in this study for the neural model. These results lead to the conclusion that we are able to predict the percentage of improvement in math that the students could achieve after playing the game, and therefore, claiming if the video game is recommendable or not for certain students.
Download

Paper Nr: 56
Title:

Cooperative Linker for the Distributed Control of the Barcelona Drinking Water Network

Authors:

Valeria Javalera-Rincon, Vicenc P. Cayuela, Bernardo M. Seix and Fernando Orduña-Cabrera

Abstract: This work shows how a Linker agent coordinates a cooperative MAS environment to seek a global optimum. The approach is applied to the Barcelona Drinking Water Network (DWN) administrated by AGBAR where the main problem was to coordinate the control of three different sectors of the network. Each part has a local controller (local agent) to solve the local water demands, but it also has to cooperate with the other agents to satisfy the water demands of the whole network. The cooperative Linker agent implemented, learns by using a Reinforcement Learning algorithm, called PlanningByExploration Behaviour with penalization (Javalera et al., 2019), to converge towards an optimal (or suboptimal) value of each of the variables that connect the local agents. For the training and simulation of the Linker agents real historical data of the Barcelona DWN provided by AGBAR were used, as well as the data to model the distributed topology of the DWN. Moreover, some results of the simulations of this approach in contrast with the results of a centralized Model Predictive Controller are depicted.
Download

Paper Nr: 58
Title:

Context-based User Activity Prediction for Mobility Planning

Authors:

Karl-Heinz Krempels, Fabian Ohler, Thomas Osterland and Christoph Terwelp

Abstract: By analyzing the individual travel characteristics of persons, it occurs that most trips are not journeys to other cities or countries but short trips, as the daily trip to work or the weekly meeting at the gym. For those trips, people know the basic conditions, as e.g., the bus driving schedule or the journey duration and it represents more effort to plan the trip beforehand, than just remember the data. But what if there is a problem, like a stalled train or a car crash on the route. Unpredictable ocurrences might be noticed too late and affect the parameters of the trip. A travelling assistant that is able to anticipate regular trips and that warns in case of problems, without requesting dedicated user input might be a solution. In this paper we consider the problem of creating an assistant based on the context information captured from a smartphone. We discuss approaches based on histogram evaluation, a Bayesian network and a multilayer perceptron that allow the prediction of locations and activities given a time and a date. These approaches are benchmarked and compared to each other to find the solution that provides the best results in prediction quality and training speed.
Download

Paper Nr: 65
Title:

Operations for Shape Transformations based on Angles

Authors:

Momo Tosue, Sosuke Moriguchi and Kazuko Takahashi

Abstract: We propose a symbolic expression for a qualitative shape of an object in the sequence of rotation angles of edges. We give a drawing algorithm for the expression based on rewriting strings and prove that we can draw a figure on a two-dimensional plane, for a consistent expression. We also refine this algorithm as an abstract rewriting system to represent shapes of figures and their changes, and prove that the system has confluence and termination.
Download

Paper Nr: 67
Title:

Kohonen Map Modification for Classification Tasks

Authors:

Jiří Jelínek

Abstract: This paper aims to present a classification model based on Kohonen maps with a modified learning mechanism and structure as well. Modification of the model structure consists of its modification for hierarchical training and recall. The change of the learning process is the transition from unsupervised to supervised learning. Experiments were performed with the modified model to verify changes in the model and compare the results with other research.
Download

Paper Nr: 80
Title:

Enhancing Siamese Networks Training with Importance Sampling

Authors:

Ajay Shrestha and Ausif Mahmood

Abstract: The accuracy of machine learning (ML) model is determined to a great extent by its training dataset. Yet the dataset optimization is often not the center of the focus to improve ML models. Datasets used in the training process can have a huge impact on the convergence of the training process and accuracy of the models. In this paper, we propose and implement importance sampling, a Monte Carlo method for variance reduction on training siamese networks to improve the accuracy of the image recognition. We demonstrate empirically that our approach can achieve improvement in training and testing errors on MNIST dataset compared to training when importance sampling is not used. Unlike standard convolution neural networks (CNN), siamese networks scale efficiently when the number of classes for image recognition increases. This paper is the first known attempt to combine importance sampling with siamese network and shows its effectiveness towards getting better accuracy.
Download

Paper Nr: 95
Title:

Multi-agent Planning System for Target Application of Earth Remote Sensing Space Systems for Solving Precision Farming Tasks

Authors:

Petr Skobelev, Vitaly Travin, Elena Simonova, Dmitry Mishutin, Vladimir Galuzin and Anastasiya Galitskaya

Abstract: The paper considers the task of constructing a schedule for target application of space remote sensing systems for solving problems of precision farming. It also proposes a method of their solution using multi-agent technology. Presented models and algorithms allow for solving the urgent problem of network interaction between heterogeneous spacecraft group devices for implementation of enlarged and detailed territory surveys, as well as adaptive planning of user requests for shooting. Results of experiments demonstrate higher efficiency of the developed algorithms on large-scale problems in comparison with methods of traditional centralized planning.
Download

Paper Nr: 104
Title:

Transfer and Extraction of the Style of Handwritten Letters using Deep Learning

Authors:

Omar Mohammed, Gérard Bailly and Damien Pellier

Abstract: How can we learn, transfer and extract handwriting styles using deep neural networks? This paper explores these questions using a deep conditioned autoencoder on the IRON-OFF handwriting data-set. We perform three experiments that systematically explore the quality of our style extraction procedure. First, We compare our model to handwriting benchmarks using multidimensional performance metrics. Second, we explore the quality of style transfer, i.e. how the model performs on new, unseen writers. In both experiments, we improve the metrics of state of the art methods by a large margin. Lastly, we analyze the latent space of our model, and we show that it separates consistently writing styles.
Download

Paper Nr: 117
Title:

Firearm Detection using Convolutional Neural Networks

Authors:

Rodrigo A. Kanehisa and Areolino A. Neto

Abstract: This papers studies the application of the YOLO algorithm to create a firearm detection system, demonstrating its effectiveness in this task. We also constructed a dataset based on the website Internet Movie Firearm Database (IMFDB) for this study. Individuals carrying firearms in public places are a strong indicator of dangerous situations. Studies show that a rapid response from law enforcement agents is the main factor in reducing the number of victims. However, a large number of cameras to be monitored leads to an overload of CCTV operators, generating fatigue and stress, consequently, loss of efficiency in surveillance. Convolutional neural networks have been shown to be efficient in the detection and identification of objects in images, having sometimes produced more accurate and consistent results than human candidates.
Download

Paper Nr: 134
Title:

SurfOpt: A New Surface Method for Optimizing the Classification of Imbalanced Dataset

Authors:

André Rodrigo da Silva, Leonardo M. Rodrigues and Luciana O. Rech

Abstract: Imbalanced classes constitute very complex machine learning classification problems, particularly if there are not many examples for training, in which case most algorithms fail to learn discriminant characteristics, and tend to completely ignore the minority class in favour of the model overall accuracy. Datasets with imbalanced classes are common in several machine learning applications, such as sales forecasting and fraud detection. Current strategies for dealing with imbalanced classes rely on manipulation of the datasets as a means of improving classification performance. Instead of optimizing classification boundaries based on some measure of distance to points, this work directly optimizes the decision surface, essentially turning a classification problem into a regression problem. We demonstrate that our approach is competitive in comparison to other classification algorithms for imbalanced classes, in addition to achieving different properties.
Download

Paper Nr: 137
Title:

Adaptive Serendipity for Recommender Systems: Let It Find You

Authors:

Miriam K. Badran, Jacques B. Abdo, Wissam Jurdi and Jacques Demerjian

Abstract: Recommender systems are nowadays widely implemented in order to predict the potential objects of interest for the user. With the wide world of the internet, these systems are necessary to limit the problem of information overload and make the user’s internet surfing a more agreeable experience. However, a very accurate recommender system creates a problem of over-personalization where there is no place for adventure and unexpected discoveries: the user will be trapped in filter bubbles and echo rooms. Serendipity is a beneficial discovery that happens by accident. Used alone, serendipity can be easily confused with randomness; this takes us back to the original problem of information overload. Hypothetically, combining accurate and serendipitous recommendations will result in a higher user satisfaction. The aim of this paper is to prove the following concept: including some serendipity at the cost of profile accuracy will result in a higher user satisfaction and is, therefore, more favourable to implement. We will be testing a first measure implementation of serendipity on an offline dataset that lacks serendipity implementation. By varying the ratio of accuracy and serendipity in the recommendation list, we will reach the optimal number of serendipitous recommendations to be included in an accurate list.
Download

Paper Nr: 144
Title:

Towards Hierarchical Probabilistic CTL Model Checking: Theoretical Foundations

Authors:

Norihiro Kamide and Yuki Yano

Abstract: This study proposes a hierarchical probabilistic computation tree logic, HpCTL, which is an extension of the standard probabilistic computation tree logic pCTL, as a theoretical basis for hierarchical probabilistic CTL model checking. Hierarchical probabilistic model checking is a new paradigm that can appropriately verify hierarchical randomized (or stochastic) systems. Furthermore, a probability-measure-independent translation from HpCTL into pCTL is defined, and a theorem for embedding HpCTL into pCTL is proved using this translation. Finally, the relative decidability of HpCTL with respect to pCTL is proved using this embedding theorem. These embedding and relative decidability results allow us to reuse the standard pCTL-based probabilistic model checking algorithms to verify hierarchical randomized systems that can be described using HpCTL.
Download

Paper Nr: 147
Title:

Changing of Participants’ Attitudes in Argument-based Negotiation

Authors:

Mare Koit

Abstract: We are modelling argument-based negotiation where the initiator is convincing the partner to do an action. The initiator is using a partner model which evaluates the hypothetical attitudes of the partner related to the action under consideration. The partner when reasoning operates with an actual model – the actual attitudes which still are hidden from the initiator. Both models are changing during negotiation as influenced by the presented arguments. The choice of an argument by a negotiation participant depends, on one hand, on the attitudes related to the action, and on the other hand, on the result of reasoning based on these attitudes. The paper studies how the participants are changing their attitudes during a dialogue. A human-human dialogue illustrates the results of the analysis of a small dialogue corpus. A limited version of the model is implemented on the computer.
Download

Paper Nr: 181
Title:

Balancing Exploitation and Exploration via Fully Probabilistic Design of Decision Policies

Authors:

Miroslav Kárný and František Hůla

Abstract: Adaptive decision making learns an environment model serving a design of a decision policy. The policy-generated actions influence both the acquired reward and the future knowledge. The optimal policy properly balances exploitation with exploration. The inherent dimensionality curse of decision making under incomplete knowledge prevents the realisation of the optimal design. This has stimulated repetitive attempts to reach this balance at least approximately. Usually, either: (a) the exploitative reward is enriched by a part reflecting the exploration quality and a feasible approximate certainty-equivalent design is made; or (b) an explorative random noise is added to the purely exploitative actions. This paper avoids the inauspicious (a) and improves (b) by employing the non-standard fully probabilistic design (FPD) of decision policies, which naturally generates random actions. Monte-Carlo experiments confirm its achieved quality. The quality stems from methodological contributions, which include: (i) an improvement of the relation between FPD and standard Markov decision processes; (ii) a design of an adaptive tuning of an FPD-parameter. The latter also suits for the tuning of the temperature in both simulated annealing and Boltzmann’s machine.
Download

Area 2 - Agents

Full Papers
Paper Nr: 8
Title:

Motivations, Classification and Model Trial of Conversational Agents for Insurance Companies

Authors:

Falko Koetter, Matthias Blohm, Monika Kochanowski, Joscha Goetzer, Daniel Graziotin and Stefan Wagner

Abstract: Advances in artificial intelligence have renewed interest in conversational agents. So-called chatbots have reached maturity for industrial applications. German insurance companies are interested in improving their customer service and digitizing their business processes. In this work we investigate the potential use of conversational agents in insurance companies by determining which classes of agents are of interest to insurance companies, finding relevant use cases and requirements, and developing a prototype for an exemplary insurance scenario. Based on this approach, we derive key findings for conversational agent implementation in insurance companies.
Download

Paper Nr: 15
Title:

Reproducing Symmetry Breaking in Exit Choice under Emergency Evacuation Situation using Response Threshold Model

Authors:

Akira Tsurushima

Abstract: When people evacuate from a room with two identical exits, it is known that these exits are often unequally used, with evacuees gathering at one of them. This inappropriate and irrational behavior sometimes results in serious loss of life. In this paper, this symmetry breaking in exit choice is discussed from the viewpoint of herding, a cognitive bias in humans during disaster evacuations. The aim of this paper is to show that the origin of symmetry breaking in exit choice is simple herd behavior, whereas many models in the literature consider the exit choice decisions either as panic or rational behavior. The evacuation decision model, based on the response threshold model in biology, is presented to reproduce human herd behavior. Simulation with the evacuation decision model shows that almost all agents gather at one exit at some frequency, despite individual agents choosing the exit randomly.
Download

Paper Nr: 69
Title:

A Voting Argumentation Framework: Considering the Reasoning behind Preferences

Authors:

Nikos Karanikolas, Pierre Bisquert and Christos Kaklamanis

Abstract: One of the most prominent ways to reach an acceptable collective decision in normal group settings is the employment of routines and methods of social choice theory. The classical social choice setting is the following: each agent involved in the decision expresses her preferences about a given set of alternatives in the form of a linear order on them. Then, the group’s aggregated decision is the outcome of the application of a voting rule to the input’s preferences. However, there are instances where social choice on its own cannot provide proper solutions. For example, there are decision problems where the outcome has to be based on the reasoning behind agents’ preferences, rather than the unjustified preferences itself. Hence, our research motivation is the practical case where agents’ rationale is needed for the decision outcome. In this paper, we explore how the agents’ rationale can be formulated inside the classical voting setting. Therefore, we propose a decision-making procedure based on argumentation and preference aggregation which permits us to explore the effect of reasoning and deliberation along with voting for the decision process. We quantify the deliberation phase by defining a new voting argumentation framework, that uses vote and generic arguments, and its acceptability semantics based on the notion of pairwise comparisons between alternatives. We prove for these semantics some theoretical results regarding well-known properties from Argumentation and Social Choice Theory.
Download

Paper Nr: 122
Title:

4-valued Logic for Agent Communication with Private/Public Information Passing

Authors:

Song Yang, Masaya Taniguchi and Satoshi Tojo

Abstract: Thus far, the agent communication has often been modeled in dynamic epistemic logic, where each agent changes his/ her belief, restricting the accessibility to possible worlds in Kripke semantics. Prior to the message passing, in general, the sender should be required to believe the contents of the message. In some occasions, however, the recipient may not believe what he/ she has heard since he/ she may not have enough background knowledge to understand it or the information may be encrypted and he/ she may not know how to decipher it. In this paper, we generalize those messages that require special knowledge as private information and formalize that the recipient does not change his/ her belief receiving such private messages. Then, we distinguish the validity of the information from the belief change of the recipient; that is, even though the communication itself is held and the information is logically contradictory to his/ her original belief, the recipient may not change his/ her belief. For this purpose, we employ 4-valued logic where each proposition is given 2 (usual true and false) times 2 (private or public information or not) truth value.
Download

Paper Nr: 162
Title:

Learning of Activity Cycle Length based on Battery Limitation in Multi-agent Continuous Cooperative Patrol Problems

Authors:

Ayumi Sugiyama, Lingying Wu and Toshiharu Sugawara

Abstract: We propose a learning method that decides the appropriate activity cycle length (ACL) according to environmental characteristics and other agents’ behavior in the (multi-agent) continuous cooperative patrol problem. With recent advances in computer and sensor technologies, agents, which are intelligent control programs running on computers and robots, obtain high autonomy so that they can operate in various fields without pre-defined knowledge. However, cooperation/coordination between agents is sophisticated and complicated to implement. We focus on the ACL which is time length from starting patrol to returning to charging base for cooperative patrol when agents like robots have batteries with limited capacity. Long ACL enable agent to visit distant location, but it requires long rest. The basic idea of our method is that if agents have long-life batteries, they can appropriately shorten the ACL by frequently recharging. Appropriate ACL depends on many elements such as environmental size, the number of agents, and workload in an environment. Therefore, we propose a method in which agents autonomously learn the appropriate ACL on the basis of the number of events detected per cycle. We experimentally indicate that our agents are able to learn appropriate ACL depending on established spatial divisional cooperation.
Download

Paper Nr: 168
Title:

Lazy Agents for Large Scale Global Optimization

Authors:

Joerg Bremer and Sebastian Lehnhoff

Abstract: Optimization problems with rugged, multi-modal Fitness landscapes, non-linear problems, and derivative-free optimization entails challenges to heuristics especially in the high-dimensional case. High-dimensionality also tightens the problem of premature convergence and leads to an exponential increase in search space size. Parallelization for acceleration often involves domain specific knowledge for data domain partition or functional or algorithmic decomposition. We extend a fully decentralized agent-based approach for a global optimization algorithm based on coordinate descent and gossiping that has no specific decomposition needs and can thus be applied to arbitrary optimization problems. Originally, the agent method suffers from likely getting stuck in high-dimensional problems. We extend a laziness mechanism that lets the agents randomly postpone actions of local optimization and achieve a better avoidance of stagnation in local optima. The extension is tested against the original method as well as against established methods. The lazy agent approach turns out to be competitive and often superior in many cases.
Download

Paper Nr: 175
Title:

A Study of Joint Policies Considering Bottlenecks and Fairness

Authors:

Toshihiro Matsui

Abstract: Multi-objective reinforcement learning has been studied as an extension of conventional reinforcement learning approaches. In the primary problem settings of multi-objective reinforcement learning, the objectives represent a trade-off between different types of utilities and costs for a single agent. Here we address a case of multiagent settings where each objective corresponds to an agent to improve bottlenecks and fairness among agents. Our major interest is how learning captures the information about the fairness with a criterion. We employ leximin-based social welfare in a single-policy, multi-objective reinforcement learning method for the joint policy of multiple agents and experimentally evaluate the proposed approach with a pursuit-problem domain.
Download

Short Papers
Paper Nr: 5
Title:

Experimental Evaluation of a Method for Simulation based Learning for a Multi-Agent System Acting in a Physical Environment

Authors:

Kun Qian, Robert W. Brehm and Lars Duggen

Abstract: A method for simulation based reinforcement learning (RL) for a multi-agent system acting in a physical environment is introduced, which is based on Multi-Agent Actor-Critic (MAAC) reinforcement learning. In the proposed method, avatar agents learn in a simulated model of the physical environment and the learned experience is then used by agents in the actual physical environment. The proposed concept is verified using a laboratory benchmark setup in which multiple agents, acting within the same environment, are required to coordinate their movement actions to prevent collisions. Three state-of-the-art algorithms for multi-agent reinforcement learning (MARL) are evaluated, with respect to their applicability for a predefined benchmark scenario. Based on simulations it is shown that the MAAC method is most applicable for implementation as it provides effective distributed learning and suits well to the concept of learning in simulated environments. Our experimental results, which compare simulated learning and task execution in a simulated environment with that of task execution in a physical environment demonstrate the feasibility of the proposed concept.
Download

Paper Nr: 39
Title:

Dynamic Path Planning with Stable Growing Neural Gas

Authors:

Carsten Hahn, Sebastian Feld, Manuel Zierl and Claudia Linnhoff-Popien

Abstract: This paper considers the problem of path planning under dynamic aspects. We propose ”Neural Gas Dynamic Path Planning” (NGDPP), a novel algorithm that continuously provides a valid path between two points inside an environment that transforms in an unpredictable manner. These transformations can occur due to both, changes in the environment’s shape and moving collision objects. The algorithm incorporates several techniques: Neural Gas, a dynamic discretization method; the A* Algorithm, a path planning algorithm for graphs; and the Potential Field method, which facilitates the avoidance of collisions. We empirically evaluate the proposed algorithm under various aspects providing performance information and guidance about situations and applications benefiting from the algorithm. The evaluation reveals that NGDPP is a solid algorithm for path planning in dynamic environments. Yet, the algorithm is based on heuristic information, i.e. a optimal result in term of the path length cannot be guaranteed.
Download

Paper Nr: 86
Title:

Constrained Coalition Formation among Heterogeneous Agents for the Multi-Agent Programming Contest

Authors:

Tabajara Krausburg and Rafael H. Bordini

Abstract: This work focuses on coalition formation among heterogeneous agents for a simulated scenario involving logistic and coordination problems. We investigate whether organising a team of agents into a number of coalitions, in which agents collaborate with each other to achieve particular goals, can increase the effectiveness of the team. We apply coalition structure generation specifically to the 2017 multi-agent programming contest, where the agents controlling various autonomous vehicles form a competing team that has to solve logistic problems simulated on the map of a real city. We experiment on three approaches with different configurations. The first uses only a task-allocation mechanism, while the other approaches use either an optimal or a heuristic coalition formation algorithm. Our results show that coalition formation can improve the performance of a participating team under some circumstances. In particular, coalition formation can indeed play an important role when we aim to balance the skills in groups of agents selected to accomplish some given set of tasks given a larger team of cooperating agents in the presence of dynamically created tasks.
Download

Paper Nr: 109
Title:

SIMSEA: A Multiagent Architecture for Fishing Activity in a Simulated Environment

Authors:

José Cascalho, Paulo Trigo, Maria J. Cruz, Armando Mendes, Eva Giacomello, Adriana Ressurreição, Tomás Dentinho and Telmo Morato

Abstract: Understanding fishermen decision-making proccess, plays a key role in predicting the impacts of the fishing activity in the marine ecosystems. Simulating fishing activity using multiagent based approaches provides tools that assist decision-makers in order to pursuit sustainable fishing activity. In this paper we present a multiagent architecture for the fishing activity where geo-referenced resources and fishing agents with different profiles are used to model and simulate the complexity of human fishing activity. A first implementation of the model (via NetLogo), along with gathered results, provides insights into the capability to build a research tool for fisheries management.
Download

Paper Nr: 130
Title:

Energy Trading in the Smart Grid: Poly-sellers Decision based on Game Theory

Authors:

Hala Alsalloum, Leila Merghem-Boulahia and Rana Rahim-Amoud

Abstract: An essential element in the smart grid is the existence of prosumers, i.e. the consumers who can also produce and sell the energy. They will become one of the stakeholders of the future grid. Their active behavior is helpful on different sides: the environmental, economical and social sides. In fact, integrating the prosumers will result in selling the surplus of energy to the grid or other consumers. However, the interactions between prosumers and the grid need to be defined in order to maximize the profit of each stakeholder. This paper proposes an energy-trading algorithm based on game theory and genetic optimization in order to optimize the satisfaction of prosumers. In our solution, buyers can afford their demands from different sellers taking into consideration the distance, the price and the amount of energy traded and needed. Simulation results indicate the effectiveness of our proposed approach in terms of minimization the total cost and maximization each prosumer satisfaction i.e. minimization the buyer’s bills and maximization the seller’s revenues.
Download

Paper Nr: 145
Title:

Combining Strengths of Optimal Multi-Agent Path Finding Algorithms

Authors:

Jiří Švancara and Roman Barták

Abstract: The problem of multi-agent path finding (MAPF) is studied in this paper. Solving MAPF optimally is a computationally hard problem and many different optimal algorithms have been designed over the years. These algorithms have good runtimes for some problem instances, while performing badly for other instances. Interestingly, these hard instances are often different across the algorithms. This leads to an idea of combining the strengths of different algorithms in such a way that an input problem instance is split into disjoint subproblems and each subproblem is solved by appropriate algorithm resulting in faster computation than using either of the algorithms for the whole instance. By manual problem decomposition we will empirically show that the above idea is viable. We will also sketch a possible future work on automated problem decomposition.
Download

Paper Nr: 152
Title:

Towards Simulated Morality Systems: Role-Playing Games as Artificial Societies

Authors:

Joan Casas-Roma, Mark J. Nelson, Joan Arnedo-Moreno, Swen E. Gaudl and Rob Saunders

Abstract: Computer role-playing games (RPGs) often include a simulated morality system as a core design element. Games’ morality systems can include both god’s eye view aspects, in which certain actions are inherently judged by the simulated world to be good or evil, as well as social simulations, in which non-player characters (NPCs) react to judgments of the player’s and each others’ activities. Games with a larger amount of social simulation have clear affinities to multi-agent systems (MAS) research on artificial societies. They differ in a number of key respects, however, due to a mixture of pragmatic game-design considerations and their typically strong embeddedness in narrative arcs, resulting in many important aspects of moral systems being represented using explicitly scripted scenarios rather than through agent-based simulations. In this position paper, we argue that these similarities and differences make RPGs a promising challenge domain for MAS research, highlighting features such as moral dilemmas situated in more organic settings than seen in game-theoretic models of social dilemmas, and heterogeneous representations of morality that use both moral calculus systems and social simulation. We illustrate some possible approaches using a case study of the morality systems in the game The Elder Scrolls IV: Oblivion.
Download

Paper Nr: 165
Title:

Extracting Primary Objects and Spatial Relations from Sentences

Authors:

Neha Baranwal, Avinash K. Singh and Suna Bensch

Abstract: In verbal human-robot interaction natural language utterances have to be grounded in visual scenes by the robot. Visual language grounding is a challenging task that includes identifying a primary object among several objects, together with the object properties and spatial relations among the objects. In this paper we focus on extracting this information from sentences only. We propose two language modelling techniques, one uses regular expressions and the other one utilizes Euclidian distance. We compare these two proposed techniques with two other techniques that utilize tree structures, namely an extended Hobb’s algorithm and an algorithm that utilizes a Stanford parse tree. A comparative analysis between all language modelling techniques shows that our proposed two approaches require less computational time than the tree-based approaches. All approaches perform good identifying the primary object and its property, but for spatial relation extraction the Stanford parse tree algorithm performs better than the other language modelling techniques. Time elapsed for the Stanford parse tree algorithm is higher than for the other techniques.
Download

Paper Nr: 166
Title:

Oscillating Mobile Neurons with Entropic Assembling

Authors:

Eugene Kagan and Shai Yona

Abstract: Functionality of neural networks is based on changing connectivity between the neurons. Usually, such changes follow certain learning procedures that define which neurons are interconnected and what is the strength of the connection. The connected neurons form the distinguished groups also known as Hebbian ensembles that can act during long time or can disintegrate into smaller groups or even into separate neurons. In the paper, we consider the mechanism of assembling / disassembling of the groups of neurons. In contrast to the traditional approaches, we set ourselves to “the neuron’s point of view” and assume that the neuron chooses the neuron to connect with following the difference between the current individual entropy and the expected entropy of the ensemble. The states of the neurons are defined by the well-known Hodgkin-Huxley model and the entropy of the neuron and the neuron’s ensemble is calculated using the Klimontovich method. The suggested model is illustrated by numerical simulations that demonstrate its close relation with the known self-organizing systems and the dynamical models of the brain activity.
Download

Paper Nr: 172
Title:

Using Multimodal Information to Enhance Addressee Detection in Multiparty Interaction

Authors:

Usman Malik, Mukesh Barange, Julien Saunier and Alexandre Pauchet

Abstract: Addressee detection is an important challenge to tackle in order to improve dialogical interactions between humans and agents. This detection, essential for turn-taking models, is a hard task in multiparty conditions. Rule based as well as statistical approaches have been explored. Statistical approaches, particularly deep learning approaches, require a huge amount of data to train. However, smart feature selection can help improve addressee detection on small datasets, particularly if multimodal information is available. In this article, we propose a statistical approach based on smart feature selection that exploits contextual and multimodal information for addressee detection. The results show that our model outperforms an existing baseline.
Download

Paper Nr: 176
Title:

Towards a Computational Approach to Emotion Elicitation in Affective Agents

Authors:

Joaquín Taverner, Emilio Vivancos and Vicente Botti

Abstract: Interest in affective computing is increasing in recent years. Different emotional approaches have been developed to incorporate emotions in multi-agent systems. However, most of these models do not offer an adequate representation of emotions. An internal representation of emotions allows to define emotions according to different affective variables. In addition, many of these approaches do not take into account factors such as culture and language when defining emotions. In this work we show the results obtained in an experiment carried out to design an affective model for a multi-agent system taking into account factors such as language and culture.
Download

Paper Nr: 189
Title:

Entropy as a Quality Measure of Correlations between n Information Sources in Multi-agent Systems

Authors:

G. Enee and J. Collonge

Abstract: Shanon’s entropy has been widely used through different Science fields, as an example, to measure the quantity of information found in a message coming from a source. In real world applications, we need to measure the quality of several crossed information sources. In the specific case of language creation within multi-agent systems, we need to measure the correlation between words and their meanings to evaluate the quality of that language. When sources of information are numerous, we are willing to make correlations between those differents sources. Considering those n sources of information are put together in a matrix having n dimensions, we propose in this paper to extend Shanon’s entropy to measure information quality in R2+ and then in Rn+.
Download

Paper Nr: 194
Title:

Mobility-oriented Agenda Planning as a Value-adding Feature for Mobility as a Service

Authors:

Felix Schwinger and Karl-Heinz Krempels

Abstract: Due to the global trend of urbanization, the current transportation networks in cities are often stressed and congested. In the coming years, this problem is going to increase further in most areas. In addition, to congestion of roads and public transportation, sustainability and emissions are becoming large factors when regarding mobility. Today’s mobility still relies on the usage of private cars in a large part, however, due to the rise of alternative travel modes and concepts such as Mobility as a Service (MaaS), the traditional mobility market may be disrupted. With the help of MaaS and the right incentives, it may be possible to shift users towards a more sustainable mobility behavior, which also relieve the stressed mobility network in cities, when more alternative mobility offers are employed. With this shift towards multimodal mobility, the complexity of searching and booking these mobility offers also rises. Users are forced to look through are large amount of alternative offers and are required to find a fitting one. In order to make this complexity more manageable, we propose a mobility-agenda planning agent that attends to the mobility needs of the users. The concept of mobility-oriented agenda planning may change how people view mobility and may provide a more holistic method of mobility planning. However, this holistic mobility planning method still has unresolved societal and technological issues that need to be addressed.
Download

Paper Nr: 3
Title:

Emergent Intelligence: A Novel Computational Intelligence Technique to Solve Problems

Authors:

Suresh Chavhan and Pallapa Venkataram

Abstract: Technological advancement and increasing globalization makes humans face many problems in day to day life, involving many possible goals and each goal is associated with multiple possible actions, each associated with many different dynamic and uncertain consequences. In real systems, the message passing mechanisms and few computational intelligence techniques (like Swarm intelligence, Multiagent System, etc.) hinder mutual cooperation and coordination of agents while solving problems in an uncertain environment, even though they are highly efficient and sophisticated. Therefore, in this paper, we propose an Emergent Intelligence technique (EIT) based problem solving. The EIT is collective intelligence of group of agents, which is an extension of multiagent system (MAS). Unlike MAS, the EIT provides independent decision making for a single task by the multiple agents with mutual coordination and cooperation. It is very useful to solve the complex and dynamic problems in uncertain environments. In this paper, we discuss EIT functioning, benefits, comparisons, and also illustration of two problems: (1) resource allocation and (2) job scheduling. Each problem is categorically analyzed and solved step by step using EIT. We measure performance of the technique by considering real time situations, and results are compared and shown the importance of EIT over MAS.
Download

Paper Nr: 22
Title:

Generation of Multiple Choice Questions Including Panoramic Information using Linked Data

Authors:

Fumika Okuhara, Yuichi Sei, Yasuyuki Tahara and Akihiko Ohsuga

Abstract: In recent years, just about all subjects require students to learn panoramic information. Because the need exists for cross-curriculum learning aimed at relating subject areas, it is useful for multiple-choice questions to include panoramic information for learners. A question including panoramic information refers to content that includes transverse related information and makes respondents grasp the whole knowledge. However, it is costly to manually generate and collect appropriate multiple-choice questions for questioners and learners. Therefore, in this research, we propose a method for the automatic generation of multiple-choice questions including panoramic information using Linked Data. Linked Data is graphical data that can link structured data, and it is used as a technology for data integration and utilization. Some attempts have been made to use Linked Data as a resource for creating teaching material, and the possibility of using Semantic Web technology in education has been verified. In this paper, we aim to realize a system for automatically generating two types of multiple-choice questions by implementing an approach to generating questions and choices. An evaluation method for the generation of questions and choices involves setting indicators for each evaluation item, such as validity and the degree of the inclusion of panoramic information.
Download

Paper Nr: 27
Title:

Asynchronous Price Stabilization Model in Networks

Authors:

Jun Kiniwa, Kensaku Kikuta and Hiroaki Sandoh

Abstract: We consider a multiagent network model consisting of nodes and edges as cities and their links to neighbors, respectively. Each network node has an agent and priced goods and the agent can buy or sell goods in the neighborhood. Though every node may not have an equal price, we can show the prices will reach an equilibrium by iterating buy and sell operations. We introduce a framework of protocols in which each buying agent makes a bid to the lowest priced goods in the neighborhood; and each selling agent selects the highest bid (if any). So far, we have just considered such a model in a synchronous environment. We, however, cannot represent the velocity of circulation of money in the synchronous system. In other words, we cannot distinguish the different speed of money movement if every operation is synchronized. Thus, we develop an asynchronous model which enables us to generalize the theory of price stabilization in networks. Finally, we execute simulation experiments and investigate the influence of network features on the velocity of money.
Download

Paper Nr: 34
Title:

Environment Engine for Situated MAS

Authors:

Halim Djerroud and Arab A. Cherif

Abstract: In the multi-agent system research community, there is general consensus that the environment is important for multi-agent systems (MAS). However, most researchers minimize the responsibilities of the environment by reducing it to inter-agent communication, or neglecting to integrate it as a main element of their MAS models, which can be sufficient depending on the focus and objectives of their work. As a consequence of these decisions, the potential of MAS is not fully exploited. In some cases, the environment is a key element that cannot be written off as inter-agent communication, as it currently is in classical MAS. In our opinion it is important to focus the MAS around the environment. Reducing the environment to inter-agent communication deprives multi-agent systems of great potential. Our point of view is that the environment is an active entity with its own processes that can change its state, regardless of the activity of its embedded agents. We propose including the environment as an entity with a set of laws. Laws can be considered rules that cannot be broken by agents. However, some researchers have been interested in integrating the environment as first class and have proposed some interesting models. Unfortunately, they are not sustained by practical applications. The aim of this paper is to contribute in two ways: first, to propose an MAS model where an environmental engine is integrated; the capabilities of this model are comparable to those of a physics engine. Second, we propose an implementation of this model and some practical cases where it can bring concrete added value.
Download

Paper Nr: 51
Title:

The Price of Anarchy: Centralized versus Distributed Resource Allocation Trade-offs

Authors:

Jinhong K. Guo, Alexander Karlovitz, Patrick Jaillet and Martin O. Hofmann

Abstract: Optimizing decision quality in large scale, distributed, resource allocation problems requires selecting the appropriate decision network architecture. Such resource allocation problems occur in distributed sensor networks, military air campaign planning, logistics networks, energy grids, etc. Optimal solutions require that demand, resource status, and allocation decisions are shared via messaging between geographically distributed, independent decision nodes. Jamming of wireless links, cyber attacks against the network, or infrastructure damage from natural disasters interfere with messaging and, thus, the quality of the allocation decisions. Our contribution described in the paper is a decentralized resource allocation architecture and algorithm that is robust to significant message loss and to uncertain demand arrival, and provides fine-grained, many-to-many combinatorial task allocation. Most importantly, it enables a conscious choice of the best level of decentralization under the expected degree of communications denial and quantifies the benefits of approximating status of peer nodes using proxy agents during temporary communications loss.
Download

Paper Nr: 71
Title:

Modeling of Emotional Influence in Multiagent System

Authors:

Jiří Jelínek

Abstract: Emotions are an integral part of human personality. That is why it is necessary to take them into account when modeling human behavior and to implement them appropriately with respect to the given objective. For simulation models, a so-called computational approach based on the emotional appraisal of the stimuli the individual is exposed to is usually used. The selection of criteria for this appraisal is not strictly given, just as the transformation of their values into the emotional space. It depends primarily on the purpose of the model and the environment in which the model exists. This paper describes a specific emotional appraisal setting for the modeling of social structures based on communication between individuals in a multi-agent environment. The experiments present simulations of several scenarios showing the development of selected model parameters over time, as well as the effect of the possible involvement of the emotional appraisal in the simulation of the authentic behavior of individuals in the network.
Download

Paper Nr: 89
Title:

Detection of Student Teacher's Intention using Multimodal Features in a Virtual Classroom

Authors:

Masato Fukuda, Hung-Hsuan Huang and Toyoaki Nishida

Abstract: The training program for high school teachers in Japan has less opportunity to practice teaching skills. As a new practice platform, we are running a project to develop a simulation platform of school environment with computer graphics animated virtual students for students’ teachers. In order to interact with virtual students and teachers, it is necessary to estimate the intention of the teacher’s behavior and utterance. However, it is difficult to detection the teacher’s intention at the classroom only by verbal information, such as whether to ask for a response or seek a response. In this paper, we propose an automatic detection model of teacher’s intention using multimodal features including linguistic, prosodic, and gestural features. For the linguistic features, we consider the models with and without lecture contents specific information. As a result, it became clear that estimating the intention of the teacher is better when using prosodic / non-verbal information together than using only verbal information. Also, the models with contents specific information perform better.
Download

Paper Nr: 93
Title:

Multi-agent Systems in Remote Sensing Image Analysis

Authors:

Peter Hofmann

Abstract: With remote sensing data and methods we gain deeper insight in many processes at the Earth’s surface. Thus, they are a valuable data source to gather geo-information of almost any kind. While the progress of remote sensing technology continues, the amount of available remote sensing data increases. Hence, besides effective strategies for data mining and image data retrieval, reliable and efficient methods of image analysis with a high degree of automation are needed in order to extract the information hidden in remote sensing data. Due to the complex nature of remote sensing data, recent methods of computer vision and image analysis do not allow a fully automatic and highly reliable analysis of remote sensing data, yet. Most of these methods are rather semi-automatic with a varying degree of automation depending on the data quality, the complexity of the image content and the information to be extracted. Thus, visual image interpretation in many cases is still seen as the most appropriate method to gather (geo-) information from remote sensing data. To increase the degree of automation, the application of multi-agent systems in remote sensing image analysis is recently under research. The paper present summarizes recent approaches and outlines their potentials.
Download

Paper Nr: 99
Title:

A Hybrid Multi-agent Architecture for Modeling in MATSim with an Alternative Scoring Strategy

Authors:

Youssef Inedjaren, Besma Zeddini, Mohamed Maachaoui and Jean-Pierre Barbot

Abstract: The development of new information and communication technologies is contributing to the emergence of a new generation of real-time services in various fields of application. In the area of intelligent transport systems, these new services also include connected vehicles that enable vehicles to collect and disseminate information, safety alerts and make driving smarter and more environmentally friendly. More and more, they concern power-assisted or fully autonomous vehicles. In this paper we propose an architecture of agents and we project it on a multi-agent transport simulator (MATSim). In order to improve the performance of the DriverAgent in the simulation an alternative approach to score the DriverAgent plans is proposed. The results show that the proposed scoring function is able to ensure that agents improve their plans at each iteration performing on the same if not better level than the current scoring function.
Download

Paper Nr: 107
Title:

Designing a Flexible Supply Chain Network with Autonomous Agents

Authors:

Takaki Matsune and Katsuhide Fujita

Abstract: This paper proposes a supply chain model that enables companies with a software agent to construct a flexible supply chain network automatically, assuming the network is composed of many competitors. Unlike the traditional supply chain model including agents, it is impossible to manage the behavior of all companies directly in the structure of the supply chain network. Each company can handle only its own strategy or planning. We also propose a simple strategy for manufacturing agents focusing on individual profit. Our experimental results demonstrate that our agents can make a supply chain network structure that produces profits in the small-to-medium-scale scenario under some sample models of various scales.
Download

Paper Nr: 115
Title:

Map Matching for SLAM with Multiple Robots in Different Moments

Authors:

André S. Oliveira, Diego O. Dantas, Doriedson M. Corrêa, Areolino A. Neto and Will M. Almeida

Abstract: This paper proposes a method for matching of occupancy grid maps in image form, in which a new way of searching for similarities is developed. The maps used were obtained via SLAM. This work uses image processing techniques to extract map features and create an alphabet, by means of features relationships, for each map. Candidates for matching points are found from the comparison between members of the alphabets generated. After the comparison, the possible matchings are verified from candidate points, applying a metric of similarity. Thus, matching points that provide greater similarity are chosen as the best current points. These operations are repeated each time that a new updated map is provided. Thus, the similarity rates of the best points of the previous iteration are updated, and the new best matching points are calculated for the current iteration; in this way, the matching points that have the highest similarity ratio among the previous iteration and the current iteration are chosen as the best current points. The results obtained are promising, since in most tests performed, it was successful in finding the correct matching between the maps. A quantitative analysis was also performed on success cases, which demonstrated the efficiency of the method and the proximity of the map matching method with single robot mapping methods.
Download

Paper Nr: 150
Title:

Analysis of the Effects of Appearances of Avatars on User's Self-evaluation of Extroversion

Authors:

Tomoko Koda and Ryosuke Oguri

Abstract: Proteus effect is known as a phenomenon in which the personality and behavior of an individual, within online virtual worlds, is changed by the appearances of their avatar. This paper reports whether Proteus effect is applicable to accompanying avatars with whom a user's avatar interacts in a virtual world. The results indicated that Proteus effect was applicable to those who perceived the avatar was their alter-ego, while those who did not was not affected by the appearance of the avatar. The results suggest the importance of considering avatar’s and agent’s appearances and user’s personality in serious games, online virtual worlds, and avatar-mediated online communications.
Download

Paper Nr: 151
Title:

An Agent-based Approach to a Temporal Headway Development Statistics in Urban Traffic using Three-phase Theory

Authors:

Maximilian Kumm and Michael Schreckenberg

Abstract: An automated vehicle is supposed to merge into the major street of a T-intersection, while disturbing the ongoing traffic as little as possible. At the same time, different requirements regarding its driving strategy have to be fulfilled with respect to safety, comfort and energy conditions. It is desirable to enable a fluent automated drive and to avoid stopping during the approach at all. We implemented an agent-based simulation using the Kerner-Klenov model in framework of the three-phase traffic theory. Using a high number of interacting vehicles leads to a multi-agent system (MAS). A normal distributed free flow parameter based on empirical traffic data is introduced and serves as an input parameter to the simulations. The simulations output yield temporal headway development-statistics, which enables a prediction of the traffic situation on the major street. This allows the automated vehicle to adjust its speed in preparation of merging into the best possible gap considering the above-mentioned requirements. Hence, taking these statistics into account helps to optimise the driving strategy of the automated vehicle.
Download