ICAART 2018 Abstracts


Area 1 - Artificial Intelligence

Full Papers
Paper Nr: 6
Title:

A Logical Approach to Extreme Opinion Diffusion

Authors:

Enzo Battistella and Laurence Cholvy

Abstract: This paper focuses on diffusion of extreme opinions among agents which influence each other. In this work, opinions are modeled as formulas of the propositional logic or equivalently, as sets of propositional interpretations. Moreover, we assume that any agent changes its opinion by merging the opinions of its influencers, from most to least influential. We propose a first definition of extreme opinions and extremism. We then consider degrees of extremism. Formal studies of these definitions are made as well as simulations.

Paper Nr: 10
Title:

Opponent Modelling in the Game of Tron using Reinforcement Learning

Authors:

Stefan J. L. Knegt, Madalina M. Drugan and Marco A. Wiering

Abstract: In this paper we propose the use of vision grids as state representation to learn to play the game Tron using neural networks and reinforcement learning. This approach speeds up learning by significantly reducing the number of unique states. Furthermore, we introduce a novel opponent modelling technique, which is used to predict the opponent’s next move. The learned model of the opponent is subsequently used in Monte-Carlo roll-outs, in which the game is simulated n-steps ahead in order to determine the expected value of conducting a certain action. Finally, we compare the performance using two different activation functions in the multi-layer perceptron, namely the sigmoid and exponential linear unit (Elu). The results show that the Elu activation function outperforms the sigmoid activation function in most cases. Furthermore, vision grids significantly increase learning speed and in most cases this also increases the agent’s performance compared to when the full grid is used as state representation. Finally, the opponent modelling technique allows the agent to learn a predictive model of the opponent’s actions, which in combination with Monte-Carlo roll-outs significantly increases the agent’s performance.

Paper Nr: 19
Title:

Constraint Networks Under Conditional Uncertainty

Authors:

Matteo Zavatteri and Luca Viganò

Abstract: Constraint Networks (CNs) are a framework to model the constraint satisfaction problem (CSP), which is the problem of finding an assignment of values to a set of variables satisfying a set of given constraints. Therefore, CSP is a satisfiability problem. When the CSP turns conditional, consistency analysis extends to finding also an assignment to these conditions such that the relevant part of the initial CN is consistent. However, CNs fail to model CSPs expressing an uncontrollable conditional part (i.e., a conditional part that cannot be decided but merely observed as it occurs). To bridge this gap, in this paper we propose constraint networks under conditional uncertainty (CNCUs), and we define weak, strong and dynamic controllability of a CNCU. We provide algorithms to check each of these types of controllability and discuss how to synthesize (dynamic) execution strategies that drive the execution of a CNCU saying which value to assign to which variable depending on how the uncontrollable part behaves. We benchmark the approach by using ZETA, a tool that we developed for CNCUs. What we propose is fully automated from analysis to simulation.

Paper Nr: 28
Title:

Actor-Critic Reinforcement Learning with Neural Networks in Continuous Games

Authors:

Gabriel Leuenberger and Marco A. Wiering

Abstract: Reinforcement learning agents with artificial neural networks have previously been shown to acquire human level dexterity in discrete video game environments where only the current state of the game and a reward are given at each time step. A harder problem than discrete environments is posed by continuous environments where the states, observations, and actions are continuous, which is what this paper focuses on. The algorithm called the Continuous Actor-Critic Learning Automaton (CACLA) is applied to a 2D aerial combat simulation environment, which consists of continuous state and action spaces. The Actor and the Critic both employ multilayer perceptrons. For our game environment it is shown: 1) The exploration of CACLA’s action space strongly improves when Gaussian noise is replaced by an Ornstein-Uhlenbeck process. 2) A novel Monte Carlo variant of CACLA is introduced which turns out to be inferior to the original CACLA. 3) From the latter new insights are obtained that lead to a novel algorithm that is a modified version of CACLA. It relies on a third multilayer perceptron to estimate the absolute error of the critic which is used to correct the learning rule of the Actor. The Corrected CACLA is able to outperform the original CACLA algorithm.

Paper Nr: 35
Title:

Deep Reinforcement Learning for Advanced Energy Management of Hybrid Electric Vehicles

Authors:

Roman Liessner, Christian Schroer, Ansgar Dietermann and Bernard Bäker

Abstract: Machine Learning seizes a substantial role in the development of future low-emission automobiles, as manufacturers are increasingly reaching limits with traditional engineering methods. Apart from autonomous driving, recent advances in reinforcement learning also offer great benefit for solving complex parameterization tasks. In this paper, deep reinforcement learning is used for the derivation of efficient operating strategies for hybrid electric vehicles. There, for achieving fuel efficient solutions, a wide range of potential driving and traffic scenarios have to be anticipated where intelligent and adaptive processes could bring significant improvements. The underlying research proves the ability of a reinforcement learning agent to learn nearlyoptimal operating strategies without any prior route-information and offers great potential for the inclusion of further variables into the optimization process.

Paper Nr: 38
Title:

Mining Substitution Rules: A Knowledge-based Approach using Dynamic Ontologies

Authors:

Rupal Sethi and B. Shekar

Abstract: Association Rule Mining has so far focused on generating and pruning positive rules using various interestingness measures. However, there are very few studies that explore the mining process of substitution rules. These studies have incorporated a limited definition of substitution, either in statistical terms or based on manager’s static knowledge. Here we attempt to provide a customer-centric model of substitution rule mining using the lens of affordance. We adopt a knowledge-based approach involving a dynamic ontology wherein objects are positioned based on the affordances they are preferred for. This contrasts with the traditional static ontology approach that highlights manager’s static knowledge base. We develop an Expected-Actual Substitution Framework to compare relatedness between items in the static and dynamic ontologies. We present Affordance-Based Substitution (ABS) algorithm to mine substitution rules based on the proposed approach. We also come up with a novel interestingness measure that enhances the quality of our substitution rules thus leading to effective knowledge discovery. Empirical analyses are performed on a real-life supermarket dataset to show the efficacy of ABS algorithm. We compare the generated rules with those generated by another substitution rule mining algorithm from the literature. Our results show that substitution rules generated through ABS algorithm capture customer perceptions that are generally missed by alternate approaches.

Paper Nr: 40
Title:

Towards Domain-independent Biases for Action Selection in Robotic Task-planning under Uncertainty

Authors:

Juan Carlos Saborío and Joachim Hertzberg

Abstract: Task-planning algorithms for robots must quickly select actions with high reward prospects despite the huge variability of their domains, and accounting for the high cost of performing the wrong action in the “real-world”. In response we propose an action selection method based on reward-shaping, for planning in (PO)MDP’s, that adds an informed action-selection bias but depends almost exclusively on a clear specification of the goal. Combined with a derived rollout policy for MCTS planners, we show promising results in relatively large domains of interest to robotics.

Paper Nr: 46
Title:

PERSEUS: A Personalization Framework for Sentiment Categorization with Recurrent Neural Network

Authors:

Siwen Guo, Sviatlana Höhn, Feiyu Xu and Christoph Schommer

Abstract: This paper introduces the personalization framework PERSEUS in order to investigate the impact of individuality in sentiment categorization by looking into the past. The existence of diversity between individuals and certain consistency in each individual is the cornerstone of the framework. We focus on relations between documents for user-sensitive predictions. Individual's lexical choices act as indicators for individuality, thus we use a concept-based system which utilizes neural networks to embed concepts and associated topics in text. Furthermore, a recurrent neural network is used to memorize the history of user's opinions, to discover user-topic dependence, and to detect implicit relations between users. PERSEUS also offers a solution for data sparsity. At the first stage, we show the benefit of inquiring a user-specified system. Improvements in performance experimented on a combined Twitter dataset are shown over generalized models. PERSEUS can be used in addition to such generalized systems to enhance the understanding of user's opinions.

Paper Nr: 47
Title:

Coalition Structure Formation using Parallel Dynamic Programming

Authors:

Samriddhi Sarkar, Pratik Sinha, Narayan Changder and Animesh Dutta

Abstract: Dynamic Programming (DP) is an effective procedure to solve many combinatorial optimization problems and optimal Coalition Structure generation (CSG) is one of them. Optimal CSG is an important combinatorial optimization problem with significant real-life applications. The current best result in terms of worst case time complexity is O(3n). So, there is a need to find speedy approaches. This paper proposes a parallel dynamic programming algorithm for optimal CSG. We performed the comparison of our algorithm with the DP for CSG problem which happens to be a sequential procedure. The theoretical, as well as the empirical results show that our proposed method works faster than its sequential counterpart. We obtain a speed-up that is almost 14 times in case of 17 agents using a 16-core machine.

Paper Nr: 50
Title:

GeoMantis: Inferring the Geographic Focus of Text using Knowledge Bases

Authors:

Christos T. Rodosthenous and Loizos Michael

Abstract: We consider the problem of identifying the geographic focus of a document. Unlike some previous work on this problem, we do not expect the document to explicitly mention the target region, making our problem one of inference or prediction, rather than one of identification. Further, we seek to tackle the problem without appealing to specialized geographic information resources like gazetteers or atlases, but employ general-purpose knowledge bases and ontologies like ConceptNet and YAGO. We propose certain natural strategies towards addressing the problem, and show that the GeoMantis system that implements these strategies outperforms an existing state-of-the-art system, when compared on documents whose target region (country, in particular) is not explicitly mentioned or is obscured. Our results give evidence that using general-purpose knowledge bases and ontologies can, in certain cases, outperform even specialized tools.

Paper Nr: 64
Title:

Deep Learning Policy Quantization

Authors:

Jos van de Wolfshaar, Marco Wiering and Lambert Schomaker

Abstract: We introduce a novel type of actor-critic approach for deep reinforcement learning which is based on learning vector quantization. We replace the softmax operator of the policy with a more general and more flexible operator that is similar to the robust soft learning vector quantization algorithm. We compare our approach to the default A3C architecture on three Atari 2600 games and a simplistic game called Catch. We show that the proposed algorithm outperforms the softmax architecture on Catch. On the Atari games, we observe a nonunanimous pattern in terms of the best performing model.

Paper Nr: 69
Title:

Hijacked Smart Devices - Methodical Foundations for Autonomous Theft Awareness based on Activity Recognition and Novelty Detection

Authors:

Martin Jänicke, Viktor Schmidt, Bernhard Sick, Sven Tomforde and Paul Lukowicz

Abstract: Personal devices such as smart phones are increasingly utilised in everyday life. Frequently, Activity Recognition is performed on these devices to estimate the current user status and trigger automated actions according to the user's needs. In this article, we focus on improving the self-awareness of such systems in terms of detecting theft: We equip devices with the capabilities to model their own user and to, e.g., alarm the legal user if an unexpected other person is carrying the device. We gathered 24hours of data in a case study with 14 persons using a Nokia N97 and trained an activity recognition system. Based on it, we developed and investigated an autonomous novelty detection system that continuously checks if the observed user behavior corresponds to the initial model, and that gives an alarm if not. Our evaluations show that the presented method is highly successful with a successful theft detection rate of over 85\% for the trained set of persons. Comparison experiments with state of the art techniques support the strong practicality of our approach.

Paper Nr: 71
Title:

TEATIME: A Formal Model of Action Tendencies in Conversational Agents

Authors:

Alya Yacoubi and Nicolas Sabouret

Abstract: This paper presents a formal model of socio-affective behaviour in a conversational agent based on the Action Tendency theory. This theory defines emotions as tendencies to perform an action. This theory allows us to implement a strong connection between emotions and speech acts during an agent-human interaction. Our model presents an agent architecture with beliefs, desires, ideals and capacities. It relies on 6 appraisal variables for the selection of different emotional strategies depending on the context of the dialogue. It also supports social regulation of emotions depending on social rules. We implemented this model in an agent architecture and we give an example of dialogue with a virtual insurance expert in the context of customer relationship management.

Paper Nr: 79
Title:

Whole Day Mobility Planning with Electric Vehicles

Authors:

Marek Cuchý, Michal Štolba and Michal Jakob

Abstract: We propose a novel and challenging variant of trip planning problems – Whole Day Mobility Planning with Electric Vehicles (WDMEV). WDMEV combines several concerns, which has been so far only considered separately, in order to realistically model the problem of planning mobility with electric vehicles (EVs). A key difference between trip planning for combustion engine cars and trip planning for EVs is the comparatively lower battery capacity and comparatively long charging times of EVs – which makes it important to carefully consider charging when planning travel. The key idea behind WDMEV is that the user can better optimize his/her mobility with EVs, if it considers the activities he/she needs to perform and the travel required to get to the locations of these activities for the whole day - rather than planning for single trips only. In this paper, we formalize the WDMEV problem and propose a solution based on a label-setting heuristic search algorithm, including several speed-ups. We evaluate the proposed algorithm on a realistic set of benchmark problems, confirming that the whole day approach reduces the time required to complete one’s day travel with EVs and that it also makes it cheaper, compared to the traditional single-trip approach.

Paper Nr: 85
Title:

Splitting-merging Clustering Algorithm for Collaborative Filtering Recommendation System

Authors:

Nabil Belacel, Guillaume Durand, Serge Leger and Cajetan Bouchard

Abstract: Collaborative filtering (CF) is a well-known and successful filtering technique that has its own limits, especially in dealing with highly sparse and large-scale data. To address this scalability issue, some researchers propose to use clustering methods like K-means that has the shortcomings of having its performances highly dependent on the manual definition of its number of clusters and on the selection of the initial centroids, which leads in case of ill-defined values to inaccurate recommendations and an increase in computation time. In this paper, we will show how the Merging and Splitting clustering algorithm can improve the performances of recommendation with reasonable computation time by comparing it with K-means based approach. Our experiment results demonstrate that the performances of our system are independent on the initial partition by considering the statistical nature of data. More specially, results in this paper provide significant evidences that the proposed splitting-merging clustering based CF is more scalable than the well-known K-means clustering based CF.

Paper Nr: 93
Title:

Speech Emotion Recognition: Methods and Cases Study

Authors:

Leila Kerkeni, Youssef Serrestou, Mohamed Mbarki, Kosai Raoof and Mohamed Ali Mahjoub

Abstract: In this paper we compare different approaches for emotions recognition task and we propose an efficient solution based on combination of these approaches. Recurrent neural network (RNN) classifier is used to classify seven emotions found in the Berlin and Spanish databases. Its performances are compared to Multivariate linear regression (MLR) and Support vector machine (SVM) classifiers. The explored features included: mel-frequency cepstrum coefficients (MFCC) and modulation spectral features (MSFs). Finally results for different combinations of the features and on different databases are compared and explained. The overall experimental results reveal that the feature combination of MFCC and MS has the highest accuracy rate on both Spanish emotional database using RNN classifier 90,05% and Berlin emotional database using MLR 82,41%.

Paper Nr: 96
Title:

Bayesian Network and Structured Random Forest Cooperative Deep Learning for Automatic Multi-label Brain Tumor Segmentation

Authors:

Samya Amiri, Mohamed Ali Mahjoub and Islem Rekik

Abstract: Brain cancer phenotyping and treatment is highly informed by radiomic analyses of medical images. Specifically, the reliability of radiomics, which refers to extracting features from the tumor image intensity, shape and texture, depends on the accuracy of the tumor boundary segmentation. Hence, developing fully-automated brain tumor segmentation methods is highly desired for processing large imaging datasets. In this work, we propose a cooperative learning framework for multi-label brain tumor segmentation, which leverages on Structured Random Forest (SRF) and Bayesian Networks (BN). Basically, we embed both strong SRF and BN classifiers into a multi-layer deep architecture, where they cooperate to better learn tumor features for our multi-label classification task. The proposed SRF-BN cooperative learning integrates two complementary merits of both classifiers. While, SRF exploits structural and contextual image information to perform classification at the pixel-level, BN represents the statistical dependencies between image components at the superpixel-level. To further improve this SRF-BN cooperative learning, we ‘deepen’ this cooperation through proposing a multi-layer framework, wherein each layer, BN inputs the original multi-modal MR images along with the probability maps generated by SRF. Through transfer learning from SRF to BN, the performance of BN improves. In turn, in the next layer, SRF will also benefit from the learning of BN through inputting the BN segmentation maps along with the original multimodal images. With the exception of the first layer, both classifiers use the output segmentation maps resulting from the previous layer, in the spirit of auto-context models. We evaluated our framework on 50 subjects with multimodal MR images (FLAIR, T1, T1-c) to segment the whole tumor, its core and enhanced tumor. Our segmentation results outperformed those of several comparison methods, including the independent (non-cooperative) learning of SRF and BN.

Paper Nr: 101
Title:

Logics and Translations for Inconsistency-tolerant Model Checking

Authors:

Norihiro Kamide and Kazuki Endo

Abstract: In this study, we develop logics and translations for inconsistency-tolerant (or paraconsistent) model checking that can be used to verify systems with inconsistencies. Paraconsistent linear-time temporal logic (pLTL) and paraconsistent computation tree logic (pCTL) are introduced, and these are extensions of standard linear-time temporal logic (LTL) and standard computation tree logic (CTL), respectively. These novel logics can be applied when handling inconsistency-tolerant temporal reasoning. These logics are also regarded as four-valued temporal logics that extend the four-valued logic of Belnap and Dunn. Translations from pLTL into LTL and pCTL into CTL are defined, and these are used to prove the theorems for embedding pLTL into LTL and pCTL into CTL. These embedding theorems allow the standard LTL- and CTL-based model checking algorithms to be used for verifying inconsistent systems that are modeled and specified by pLTL and pCTL. A new illustrative example for inconsistency-tolerant model checking is also presented on the basis of the proposed logics and translations.

Paper Nr: 113
Title:

Concept Similarity under the Agent’s Preferences for the Description Logic FL0 with Unfoldable TBox

Authors:

Teeradaj Racharak and Satoshi Tojo

Abstract: Concept similarity refers to human judgment of a degree to which a pair of concepts is similar. Computational techniques attempting to imitate such judgment are called concept similarity measures. In Description Logics (DLs), we could regard them as a generalization of the classical reasoning problem of equivalence. That is, any two concepts are equivalent if and only if their similarity degree is one. When two concepts are not equivalent, the level of similarity varies depending not only on the objective factors (e.g. the structure of concept descriptions) but also on the subjective factors (i.e. the agent’s preferences). The recently introduced notion called preference profile identified a collection of preferential elements in which any developments for concept similarity measure should consider. In this paper, we briefly review approaches of identifying the subsumption degree between FL0 concept descriptions and exemplify how one can adopt the viewpoint of preference profile toward the development of concept similarity measure under the agent’s preferences in FL0. Finally, we investigate several properties of the developed measure and discuss future directions.

Paper Nr: 118
Title:

Social Emotion Mining Techniques for Facebook Posts Reaction Prediction

Authors:

Florian Krebs, Bruno Lubascher, Tobias Moers, Pieter Schaap and Gerasimos Spanakis

Abstract: As of February 2016 Facebook allows users to express their experienced emotions about a post by using five so-called ‘reactions’. This research paper proposes and evaluates alternative methods for predicting these reactions to user posts on public pages of firms/companies (like supermarket chains). For this purpose, we collected posts (and their reactions) from Facebook pages of large supermarket chains and constructed a dataset which is available for other researches. In order to predict the distribution of reactions of a new post, neural network architectures (convolutional and recurrent neural networks) were tested using pretrained word embeddings. Results of the neural networks were improved by introducing a bootstrapping approach for sentiment and emotion mining on the comments for each post. The final model (a combination of neural network and a baseline emotion miner) is able to predict the reaction distribution on Facebook posts with a mean squared error (or misclassification rate) of 0.135.

Paper Nr: 125
Title:

On the Effects of Team Size and Communication Load on the Performance in Exploration Games

Authors:

Chris Rozemuller, Mark Neerincx and Koen Hindriks

Abstract: Exploration games are games where agents (or robots) need to search resources and retrieve these resources. In principle, performance in such games can be improved either by adding more agents or by exchanging more messages. However, both measures are not free of cost and it is important to be able to assess the trade-off between these costs and the potential performance gain. The focus of this paper is on improving our understanding of the performance gain that can be achieved either by adding more agents or by increasing the communication load. Performance gain moreover is studied by taking several other important factors into account such as environment topology and size, resource-redundancy, and task size. Our results suggest that there does not exist a decision function that dominates all other decision functions, i.e. is optimal for all conditions. Instead we find that (i) for different team sizes and communication strategies different agent decision functions perform optimal, and that (ii) optimality of decision functions also depends on environment and task parameters. We also find that it pays off to optimize for environment topologies.

Paper Nr: 130
Title:

Gamma-star Reduction in the Type-theory of Acyclic Algorithms

Authors:

Roussanka Loukanova

Abstract: The paper extends a higher-order type theory of acyclic algorithms by adding a reduction rule, which results in a stronger reduction calculus. The new reduction calculus determines a strong algorithmic equivalence between formal terms. It is very useful for simplifying terms, by eliminating sub-terms having superfluous lambda abstraction and corresponding spurious functional applications.

Paper Nr: 132
Title:

Experience Filtering for Robot Navigation using Deep Reinforcement Learning

Authors:

Phong Nguyen, Takayuki Akiyama and Hiroki Ohashi

Abstract: We propose a stochastic method of storing a new experience into replay memory to increase the performance of the Deep Q-learning (DQL) algorithm, especially under the condition of a small memory. The conventional standard DQL method with the Prioritized Experience Replay method attempts to use experiences in the replay memory for improving learning efficiency; however, it does not guarantee the diversity of experience in the replay memory. Our method calculates the similarity of a new experience with other existing experiences in the memory based on a distance function and determines whether to store this new experience stochastically. This method leads to the improvement in experience diversity in the replay memory and better utilization of rare experiences during the training process. In an experiment to train a moving robot, our proposed method improved the performance of the standard DQL algorithm with a memory buffer of less than 10,000 stored experiences.

Paper Nr: 143
Title:

Predicting Temperament using Keirsey’s Model for Portuguese Twitter Data

Authors:

Cristina Fátima Claro, Ana Carolina E. S. Lima and Leandro N. de Castro

Abstract: Temperament is a set of innate tendencies of the mind related with the processes of perceiving, analyzing and decision making. The purpose of this paper is to predict the user's temperament based on Portuguese tweets and following Keirsey's model, which classifies the temperament into artisan, guardian, idealist and rational. The proposed methodology uses a Portuguese version of LIWC, which is a dictionary of words, to analyze the context of words, and supervised learning using the KNN, SVM and Random Forest algorithms for train-ing the classifiers. The resultant average accuracy obtained was 88.37% for the artisan temperament, 86.92% for the guardian, 55.61% for the idealist, and 69.09% for the rational. By using binary classifiers the average accuracy was 90.93% for the artisan temperament, 88.98% for the guardian, 51.98% for the idealist and 71.42% for the Rational.

Paper Nr: 144
Title:

A Constraint Solving Web Service for Recognizing Historical Japanese KANA Texts

Authors:

Kazuki Sando, Tetsuya Suzuki and Akira Aiba

Abstract: One of the first steps for researching Japanese classical literature is reading Japanese historical manuscripts. However the reading process is not easy, and time-consuming since a set of characters used in those manuscripts contain different characters from those currently used. There have been several attempts to read Japanese historical manuscripts. We proposed a framework to assist the human process for reading Japanese historical manuscript. It formulates the process as a constraint satisfaction problem, and a constraint solver in the whole system, was experimentally implemented as a UNIX command. In this paper, we added a Web service layer to the solver to realize loose coupling between the solver and the other subsystems. Thanks to the loose coupling, any programming language can be used for implementation of other parts of the whole system. In addition, the constraint solving Web service can be public through the Internet. We experimentally confirmed the solver as a Web service is faster than the that as a UNIX command if both the solver and a client are connected to a same local area network.

Paper Nr: 152
Title:

The Mapping Distance – a Generalization of the Edit Distance – and its Application to Trees

Authors:

Kilho Shin and Taro Niiyama

Abstract: The edit distances has been widely used as an effective method to analyze similarity of compound data, which consist of multiple components, such as strings, trees and graphs. For example, the Levenshtein distance for strings is known to be effective to analyze DNA and proteins, and the Ta¨ı distance and its variations are attracting wide attention of researchers who study tree-type data such as glycan, HTML-DOM-trees, parse trees of natural language processing and so on. The problem that we recognize here is that the way of engineering new edit distances was ad-hoc and lacked a unified view. To solve the problem, we introduce the concept of the mapping distance. The mapping distance framework can provide a unified view over various distance measures for compound data focusing on partial one-to-one mappings between data. These partial one-to-one mappings are a generalization of what are known as traces in the legacy study of edit distances. This is a clear contrast to the legacy edit distance framework, which define distances between compound data through edit operations and edit paths. Our framework enables us to design new distance measures consistently, and also, various distance measures can be described using a small number of parameters. In fact, in this paper, we take rooted trees as an example and introduce three independent dimensions to parameterize mapping distance measures. As a result, we define 16 mapping distance measures, 13 of which are novel. In experiments, we discover that some novel measures outperform the others including the legacy edit distances in accuracy when used with the k-NN classifier.

Paper Nr: 154
Title:

Estimation of Reward Function Maximizing Learning Efficiency in Inverse Reinforcement Learning

Authors:

Yuki Kitazato and Sachiyo Arai

Abstract: Inverse Reinforcement Learning (IRL) is a promising framework for estimating a reward function given the behavior of an expert.However, the IRL problem is ill-posed because infinitely many reward functions can be consistent with the expert’s observed behavior. To resolve this issue, IRL algorithms have been proposed to determine alternative choices of the reward function that reproduce the behavior of the expert, but these algorithms do not consider the learning efficiency. In this paper, we propose a new formulation and algorithm for IRL to estimate the reward function that maximizes the learning efficiency. This new formulation is an extension of an existing IRL algorithm, and we introduce a genetic algorithm approach to solve the new reward function. We show the effectiveness of our approach by comparing the performance of our proposed method against existing algorithms.

Short Papers
Paper Nr: 5
Title:

Fuzzy Contagion Cascades in Financial Networks

Authors:

Giuseppe De Marco, Chiara Donnini, Federica Gioia and Francesca Perla

Abstract: Previous literature shows that financial networks are sometimes described by fuzzy data. This paper extends classical models of financial contagion to the framework of fuzzy financial networks. The degree of default of a bank in the network consists in a (real valued) measure of the fuzzy default and it is computed as a fixed point for the dynamics of a modified ”fictitious default algorithm”. Finally, the algorithm is implemented in MATLAB and tested numerically on a real data set.

Paper Nr: 11
Title:

Concise Finite-Domain Representations for Factored MA-PDDL Planning Tasks

Authors:

Daniel Fišer and Antonín Komenda

Abstract: Planning tasks for the distributed multi-agent planning in deterministic environments are described in highly expressive, but lifted, languages, similar to classical planning. On the one hand, these languages allow for the compact representation of exponentially large planning problems. On the other hand, the solvers using such languages need efficient grounding methods to translate the high-level description to a low-level representation using facts or atomic values. Although there exist ad-hoc implementations of the grounding for the multi-agent planning, there is no general scheme usable by all multi-agent planners. In this work, we propose such a scheme combining centralized processes of the grounding and the inference of mutex groups. Both processes are needed for the translation of planning tasks from the Multi-agent Planning Description Language (MA-PDDL) to the finite domain representation. We experimentally show a space reduction of the multi-agent finite domain representation in contrast to the binary representation on the common benchmark set.

Paper Nr: 13
Title:

ICM: An Intuitive Model Independent and Accurate Certainty Measure for Machine Learning

Authors:

Jasper van der Waa, Jurriaan van Diggelen, Mark Neerincx and Stephan Raaijmakers

Abstract: End-users of machine learning-based systems benefit from measures that quantify the trustworthiness of the underlying models. Measures like accuracy provide for a general sense of model performance, but offer no detailed information on specific model outputs. Probabilistic outputs, on the other hand, express such details, but they are not available for all types of machine learning, and can be heavily influenced by bias and lack of representative training data. Further, they are often difficult to understand for non-experts. This study proposes an intuitive certainty measure (ICM) that produces an accurate estimate of how certain a machine learning model is for a specific output, based on errors it made in the past. It is designed to be easily explainable to non-experts and to act in a predictable, reproducible way. ICM was tested on four synthetic tasks solved by support vector machines, and a real-world task solved by a deep neural network. Our results show that ICM is both more accurate and intuitive than related approaches. Moreover, ICM is neutral with respect to the chosen machine learning model, making it widely applicable.

Paper Nr: 14
Title:

Unfolding Ensemble Training Sets for Improved Support Vector Decoders in Energy Management

Authors:

Joerg Bremer and Sebastian Lehnhoff

Abstract: Smart grid control demands delegation of liabilities to distributed, rather small energy resources in contrast to todays large control power units. Distributed energy scheduling constitutes a complex task for optimization algorithms regarding the underlying high-dimensional, multimodal and nonlinear problem structure. Additionally, the necessity for abstraction from individual capabilities is given while integrating energy units into a general optimization model. For predictive scheduling with high penetration of renewable energy resources, agent-based approaches using classifier-based decoders for modeling individual flexibilities have shown good performance. On the other hand, such decoder-based methods are currently designed for single entities and not able to cope with ensembles of energy resources. Combining training sets randomly sampled from individually modeled energy units, results in folded distributions with unfavorable properties for training a decoder. Nevertheless, this happens to be a quite frequent use case, e. g. when a hotel, a small business, a school or similar with an ensemble of co-generation, heat pump, solar power, and controllable consumers wants to take part in decentralized predictive scheduling. We use a Simulated Annealing approach to correct the unsuitable distribution of instances in the aggregated ensemble training set prior to deriving a flexibility model. Feasibility is ensured by integrating individual flexibility models of the respective energy units as boundary penalty while the mutation drives instances from the training set through the feasible region of the energy ensemble. Applicability is demonstrated by several simulations using established models for energy unit simulation.

Paper Nr: 16
Title:

Lowest Unique Bid Auctions with Resubmission Opportunities

Authors:

Yida Xu and Hamidou Tembine

Abstract: The recent online platforms propose multiple items for bidding. The state of the art, however, is limited to the analysis of one item auction. In this paper we study multi-item lowest unique bid auctions (LUBA) in discrete bid spaces under budget constraints. We show the existence of mixed Bayes-Nash equilibria for an arbitrary number of bidders and items. The equilibrium is explicitly computed in two bidder setup with resubmission possibilities. In the general setting we propose a distributed strategic learning algorithm to approximate equilibria. Computer simulations indicate that the error quickly decays in few number of steps by means of speedup techniques. When the number of bidders per item follows a Poisson distribution, it is shown that the seller can get a non-negligible revenue on several items, and hence making a partial revelation of the true value of the items.

Paper Nr: 25
Title:

A Multi-Objective, Risk-based Approach for Selecting Software Requirements

Authors:

Aruan G. Amaral and Gledson Elias

Abstract: In iterative and incremental development approaches, there is great interest in delivering system releases on-budget, but raising stakeholders’ satisfaction as much as possible. In the field of Search Based Software Engineering (SBSE), such a problem is known as the Next Release Problem (NRP), which is handled in existing proposals by reformulating the requirements selection process as an optimization problem solved by metaheuristics, providing a set of recommendations with the highest customers’ satisfactions as well as the lowest development costs. Despite their contributions, most of current proposals do not address software risks, which represent a key aspect that can deeply impact on project cost and stakeholders’ satisfaction. In such a direction, this paper proposes a multi-objective, risk-based approach for the NRP problem, in which a risk analysis is incorporated to estimate the impact of software risks in development cost and stakeholders’ satisfaction. Experimental results reveal the efficiency and practical applicability of the proposed approach.

Paper Nr: 26
Title:

Multidimensional Representations for the Gesture Phase Segmentation Problem - An Exploratory Study using Multilayer Perceptrons

Authors:

Ricardo A. Feitosa, Jallysson M. Rocha, Clodoaldo A. M. Lima and Sarajane M. Peres

Abstract: Gesture analysis systems have been attracting a good deal of attention because of the improvements they have made to the interaction between humans, humans and machines, and humans and their environment. In this interaction, natural gesticulation can be regarded as a part of the linguistic system underlying the communication, and the whole information system that seeks to make use of this kind of interaction for making decisions, should be able to “interpret” it. This can be carried out through strategies for gesture phase segmentation. The establishment of an efficient data representation for gestures is a critical issue when undertaking this task. The chosen representation, as well as the way it is combined with analytical techniques, may or may not support the solution that is found. In this study, different forms of representation for gestures are applied to a Multilayer Perceptron to create a suitable environment for detecting the more discriminative representations. The results obtained in this study showed that spatial and temporal characteristics must be combined to build discriminatory gesture representation, for the context of gesture phase segmentation.

Paper Nr: 30
Title:

Dynamic Repairing A*: a Plan-Repairing Algorithm for Dynamic Domains

Authors:

Filippos Gouidis, Theodore Patkos, Giorgos Flouris and Dimitris Plexousakis

Abstract: Re-planning is a special case of planning which arises when already produced plans become invalidated before their completion. In this work we investigate the conditions under which plan repairing is more efficient than re-planning from scratch. We present a new plan-repairing algorithm, Dynamic Repairing A* (DRA*) and we compare its performance against A* in a number of different re-planning scenarios. The experimental results indicate that if the percentage of the plan that has been already executed is less than 40% to 50% and the changes in the environment are small or moderate, DRA* outperforms A* in terms of speed by a factor of 10% to 80% in the majority of the cases.

Paper Nr: 39
Title:

Saliency based Adjective Noun Pair Detection System

Authors:

Marco Stricker, Syed Saqib Bukhari, Damian Borth and Andreas Dengel

Abstract: This paper investigates if it is possible to increase the accuracy of Convolutional Neural Networks trained on Adjective Noun Concepts with the help of saliency models. Although image classification reaches high accuracy rates, the same level of accuracy is not reached for Adjective Noun Pairs, due to multiple problems. Several benefits can be gained through understanding Adjective Noun Pairs, like automatically tagging large image databases and understanding the sentiment of these images. This knowledge can be used for e.g. a better advertisement system. In order to improve such a sentiment classification system a previous work focused on searching saliency methods that can reproduce the human gaze on Adjective Noun Pairs and found out that “Graph-Based Visual Saliency” belonged to the best for this problem. Utilizing these results we used the “Graph-Based Visual Saliency” method on a big dataset of Adjective Noun Pairs and incorporated these saliency data in the training phase of the Convolutional Neural Network. We tried out three different approaches to incorporate this information in three different cases of Adjective Noun Pair combinations. These cases either share a common adjective or a common noun or are completely different. Our results showed only slight improvements which were not significantly better besides for one technique in one case.

Paper Nr: 44
Title:

Machine Floriography: Sentiment-inspired Flower Predictions over Gated Recurrent Neural Networks

Authors:

Avi Bleiweiss

Abstract: The design of a flower bouquet often comprises a manual step of plant selection that follows an artistic style arrangement. Floral choices for a collection are typically founded on visual aesthetic principles that include shape, line, and color of petals. In this paper, we propose a novel framework that instead classifies sentences that describe sentiments and emotions typically conveyed by flowers, and predicts the bouquet content implicitly. Our work exploits the figurative Language of Flowers that formalizes an expandable list of translation records, each mapping a short-text sentiment sequence to a unique flower type we identify with the bouquet center-of-interest. Records are represented as word embeddings we feed into a gated recurrent neural-network, and a discriminative decoder follows to maximize the score of the lead flower and rank complementary flower types based on their posterior probabilities. Already normalized, these scores directly shape the mix weights in the final arrangement and support our intuition of a naturally formed bouquet. Our quantitative evaluation reviews both stand-alone and baseline comparative results.

Paper Nr: 49
Title:

SentiCite - An Approach for Publication Sentiment Analysis

Authors:

Dominique Mercier, Akansha Bhardwaj, Andreas Dengel and Sheraz Ahmed

Abstract: With the rapid growth in the number of scientific publications, year after year, it is becoming increasingly difficult to identify quality authoritative work on a single topic. Though there is an availability of scientometric measures which promise to offer a solution to this problem, these measures are mostly quantitative and rely, for instance, only on the number of times an article is cited. With this approach, it becomes irrelevant if an article is cited 10 times in a positive, negative or neutral way. In this context, it is quite important to study the qualitative aspect of a citation to understand its significance. This paper presents a novel system for sentiment analysis of citations in scientific documents (SentiCite) and is also capable of detecting nature of citations by targeting the motivation behind a citation, e.g., reference to a dataset, reading reference. Furthermore, the paper also presents two datasets (SentiCiteDB and IntentCiteDB) containing about 2,600 citations with their ground truth for sentiment and nature of citation. SentiCite along with other state-of-the-art methods for sentiment analysis are evaluated on the presented datasets. Evaluation results reveal that SentiCite outperforms state-of-the-art methods for sentiment analysis in scientific publications by achieving a F1-measure of 0.71.

Paper Nr: 55
Title:

Mapping Data Sets to Concepts using Machine Learning and a Knowledge based Approach

Authors:

Andreas Bunte, Peng Li and Oliver Niggemann

Abstract: Machine learning techniques have a huge potential to take some tasks of humans, e.g. anomaly detection or predictive maintenance, and thus support operators of cyber physical systems (CPSs). One challenge is to communicate algorithms results to machines or humans, because they are on a sub-symbolical level and thus hard to interpret. To simplify the communication and thereby the usage of the results, they have to be transferred to a symbolic representation. Today, the transformation is typically static which does not satisfy the needs for fast changing CPSs and prohibit the usage of the full machine learning potential. This work introduces a knowledge based approach of an automatic mapping between the sub-symbolic results of algorithms and their symbolic representation. Clustering is used to detect groups of similar data points which are interpreted as concepts. The information of clusters are extracted and further classified with the help of an ontology which infers the current operational state. Data from wind turbines is used to evaluate the approach. The achieved results are promising, the system can identify its operational state without an explicit mapping.

Paper Nr: 57
Title:

Towards a Digital Personal Trainer for Health Clubs - Sport Exercise Recognition Using Personalized Models and Deep Learning

Authors:

Sebastian Baumbach, Arun Bhatt, Sheraz Ahmed and Andreas Dengel

Abstract: Human activity recognition has emerged as an active research area in recent years. With the advancement in mobile and wearable devices, various sensors are ubiquitous and widely available gathering data a broad spectrum of peoples’ daily life activities. Research studies thoroughly assessed lifestyle activities and are increasingly concentrated on a variety of sport exercises. In this paper, we examine nine sport and fitness exercises commonly conducted with sport equipments in gym, such as abdominal exercise and lat pull. We collected sensor data of 23 participants for these activities, for which smartphones and smartwatches were used. Traditional machine learning and deep learning algorithms were applied in these experiments in order to assess their performance on our dataset. Linear SVM and Naive Bayes with Gaussian kernel performs best with an accuracy of 80 %, whereas deep learning models outperform these machine learning techniques with an accuracy of 92 %.

Paper Nr: 59
Title:

Factors Affecting Accuracy in Image Translation based on Generative Adversarial Network

Authors:

Fumiya Yamashita, Ryohei Orihara, Yuichi Sei, Yasuyuki Tahara and Akihiko Ohsuga

Abstract: With the development of deep learning, image translation has made it possible to output more realistic and highly accurate images. Especially, with the advent of Generative Adversarial Network (GAN), it became possible to perform general purpose learning in various image translation tasks such as “drawings to paintings”, “male to female” and “day to night”. In recent works, several models have been proposed that can do unsupervised learning which does not require an explicit pair of source domain image and target domain image, which is conventionally required for image translation. Two models called “CycleGAN” and “DiscoGAN” have appeared as state-of-the-art models in unsupervised learning-based image translation and succeeded in creating more realistic and highly accurate images. These models share the same network architecture, although there are differences in detailed parameter settings and learning algorithms. (in this paper we will collectively refer to them as “learning techniques”) Both models can do similar translation tasks, but it turned out that there is a large difference in translation accuracy between particular image domains. In this study, we analyzed differences in learning techniques of these models and investigated which learning techniques affect translation accuracy. As a result, it was found that the difference in the size of the feature map, which is the input for the image creation, affects the accuracy.

Paper Nr: 60
Title:

Do Professional Football Players Follow the Optimal Strategies in Penalty Shootout?

Authors:

Takaya Koizumi, Ryohei Orihara, Yuichi Sei, Yasuyuki Tahara and Akihiko Ohsuga

Abstract: Do people and companies choose optimal strategies under various situations? If not so, why not? Pursuing the reason for this helps to understand individuals and companies. In addition, game theory has been heavily involved in the understanding of sports, economics and other social sciences. In this study, we focused on football’s penalty shootout where data is relatively easy to collect, mixed strategy can be applied and making a pay-off matrix considering our own probability is possible. Using the pay-off matrix, we obtained optimal strategy of the kicker in the penalty shootout and revealed the gap between the optimal strategy and actual action taken by players. We compared the probability distribution for each data attribute in the dataset in order to obtain the cause of the gap. We use 100 professional penalty shootouts (total 1032 kicks) which were collected from internet video site during the period from 2001-2016. Experimental results showed that there was a gap between the optimal strategy and the actual action taken by players and that it also suggested the position and team attributes and temporary scores of the shootout and kicking order involved in the gap. Considering them we made the hypothesis and estimated the cause of the gap. We hope this method can apply to other fields than sports.

Paper Nr: 61
Title:

Detecting Dutch Political Tweets: A Classifier based on Voting System using Supervised Learning

Authors:

Eric Fernandes de Mello Araújo and Dave Ebbelaar

Abstract: The task of classifying political tweets has been shown to be very difficult, with controversial results in many works and with non-replicable methods. Most of the works with this goal use rule-based methods to identify political tweets. We propose here two methods, being one rule-based approach, which has an accuracy of 62%, and a supervised learning approach, which went up to 97% of accuracy in the task of distinguishing political and non-political tweets in a corpus of 2.881 Dutch tweets. Here we show that for a data base of Dutch tweets, we can outperform the rule-based method by combining many different supervised learning methods.

Paper Nr: 66
Title:

Hierarchical Reinforcement Learning for Real-Time Strategy Games

Authors:

Remi Niel, Jasper Krebbers, Madalina M. Drugan and Marco A. Wiering

Abstract: Real-Time Strategy (RTS) games can be abstracted to resource allocation applicable in many fields and industries. We consider a simplified custom RTS game focused on mid-level combat using reinforcement learning (RL) algorithms. There are a number of contributions to game playing with RL in this paper. First, we combine hierarchical RL with a multi-layer perceptron (MLP) that receives higher-order inputs for increased learning speed and performance. Second, we compare Q-learning against Monte Carlo learning as reinforcement learning algorithms. Third, because the teams in the RTS game are multi-agent systems, we examine two different methods for assigning rewards to agents. Experiments are performed against two different fixed opponents. The results show that the combination of Q-learning and individual rewards yields the highest win-rate against the different opponents, and is able to defeat the opponent within 26 training games.

Paper Nr: 70
Title:

Hierarchical Model for Zero-shot Activity Recognition using Wearable Sensors

Authors:

Mohammad Al-Naser, Hiroki Ohashi, Sheraz Ahmed, Katsuyuki Nakamura, Takayuki Akiyama, Takuto Sato, Phong Nguyen and Andreas Dengel

Abstract: We present a hierarchical framework for zero-shot human-activity recognition that recognizes unseen activities by the combinations of preliminarily learned basic actions and involved objects. The presented framework consists of gaze-guided object recognition module, myo-armband based action recognition module, and the activity recognition module, which combines results from both action and object module to detect complex activities. Both object and action recognition modules are based on deep neural network. Unlike conventional models, the proposed framework does not need retraining for recognition of an unseen activity, if the activity can be represented by a combination of the predefined basic actions and objects. This framework brings competitive advantage to industry in terms of the service-deployment cost. The experimental results showed that the proposed model could recognize three types of activities with precision of 77% and recall rate of 82%, which is comparable to a baseline method based on supervised learning.

Paper Nr: 72
Title:

Sentence and Word Embedding Employed in Open Question-Answering

Authors:

Marek Medveď and Aleš Horák

Abstract: The Automatic Question Answering, or AQA, system is a representative of open domain QA systems, where the answer selection process leans on syntactic and semantic similarities between the question and the answering text snippets. Such approach is specifically oriented to languages with fine grained syntactic and morphologic features that help to guide the correct QA match. In this paper, we present the latest results of the AQA system with new word embedding criteria implementation. All AQA processing steps (question processing, answer selection and answer extraction) are syntax-based with advanced scoring obtained by a combination of several similarity criteria (TF-IDF, tree distance, ...). Adding the word embedding parameters helped to resolve the QA match in cases, where the answer is expressed by semantically near equivalents. We describe the design and implementation of the whole QA process and provide a new evaluation of the AQA system with the word embedding criteria measured with an expanded version of Simple Question-Answering Database, or SQAD, with more than 3,000 question-answer pairs extracted from the Czech Wikipedia.

Paper Nr: 88
Title:

Dynamic Pricing Strategy for Electromobility using Markov Decision Processes

Authors:

Jan Mrkos, Antonín Komenda and Michal Jakob

Abstract: Efficient allocation of charging capacity to electric vehicle (EV) users is a key prerequisite for large-scale adaption of electric vehicles. Dynamic pricing represents a flexible framework for balancing the supply and demand for limited resources. In this paper, we show how dynamic pricing can be employed for allocation of EV charging capacity. Our approach uses Markov Decision Process (MDP) to implement demand-response pricing which can take into account both revenue maximization at the side of the charging station provider and the minimization of cost of charging on the side of the EV driver. We experimentally evaluate our method on a real-world data set. We compare our dynamic pricing method with the flat rate time-of-use pricing that is used today by most paid charging stations and show significant benefits of dynamically allocating charging station capacity through dynamic pricing.

Paper Nr: 102
Title:

Improving Text Classification with Vectors of Reduced Precision

Authors:

Krzysztof Wróbel, Maciej Wielgosz, Marcin Pietron, Michal Karwatowski, Jerzy Duda and Aleksander Smywinski-Pohl

Abstract: This paper presents the analysis of the impact of a floating-point number precision reduction on the quality of text classification. The precision reduction of the vectors representing the data (e.g. TF–IDF representation in our case) allows for a decrease of computing time and memory footprint on dedicated hardware platforms. The impact of precision reduction on the classification quality was performed on 5 corpora, using 4 different classifiers. Also, dimensionality reduction was taken into account. Results indicate that the precision reduction improves classification accuracy for most cases (up to 25% of error reduction). In general, the reduction from 64 to 4 bits gives the best scores and ensures that the results will not be worse than with the full floating-point representation.

Paper Nr: 103
Title:

Medical Decision Support Tool from a Fuzzy-Rules Driven Bayesian Network

Authors:

Vasilios Zarikas, Elpiniki Papageorgiou, Damira Pernebayeva and Nurislam Tursynbek

Abstract: The task of carrying out an effective and efficient decision on medical domain is a complex one, since a lot of uncertainty and vagueness is involved. Fuzzy logic and probabilistic methods for handling uncertain and imprecise data both provide an advance towards the goal of constructing an intelligent decision support system (DSS) for medical diagnosis and therapy. This work reports on a successfully developed DSS concerning pneumonia disease. A detailed and clear description of the reasoning behind the core decision making module of the DSS, is included, depicting the proposed methodological issues. The results have shown that the suggested methodology for constructing bayesian networks (BNs) from fuzzy rules gives a front-end decision about the severity of pulmonary infections, providing similar results to those obtained with physicians’ intuition.

Paper Nr: 105
Title:

Predicting Read- and Write-Operation Availabilities of Quorum Protocols based on Graph Properties

Authors:

Robert Schadek, Oliver Kramer and Oliver Theel

Abstract: Highly available services can be implemented by means of quorum protocols. Unfortunately, using real-world physical networks as underlying communication medium for quorum protocols turns out to be difficult, since efficient quorum protocols often depend on a particular graph structure imposed on the replicas managed by it. Mapping the replicas of the quorum protocol to the vertices of the real-world physical network usually decreases the availability of the operation provided by the quorum protocol. Therefore, finding mappings with little decrease in operation availability is the desired goal. The mapping with the smallest decrease in operation availability can be found by iterating all mappings. This approach has a runtime complexity of O(N!) where N is the number of vertices in the graph structure. Finding the optimal mapping with this approach, therefore, quickly becomes unfeasible. We present, an approach to predict the operation availability of the best mapping based on properties like e. g. degree or betweenness centrality. This prediction can then be used to decide whether it is worth to execute the O(N!) algorithm to find the best possible mapping. We test this new approach by cross-validating its predictions of the operation availability with the operation availability of the best mapping.

Paper Nr: 110
Title:

Possibilistic Morphological Disambiguation of Structured Hadiths Arabic Texts using Semantic Knowledge

Authors:

Raja Ayed, Bilel Elayeb and Narjès Bellamine Ben Saoud

Abstract: We propose, in this paper, a possibilistic morphological approach to disambiguate hadiths Arabic texts using semantic knowledge. The disambiguation is considered as a classification problem. The possibilistic approach uses vocalized texts to train a possibilistic classifier in order to classify non-vocalized texts as they are more ambiguous. Morphological attributes are used for training and test. Hadiths are structured in XML format that provides semantic information. We enlarge the classification attributes’ set by adding semantic attributes extracted from the hadiths structure. We prove that the possibilistic approach gives the best rates using AlKhalil analyzer to prepare the training and the test sets. Our proposed possibilistic approach enhances disambiguation rates of Arabic hadiths’ texts when it includes semantic knowledge.

Paper Nr: 114
Title:

A Lock-free Algorithm for Parallel MCTS

Authors:

S. Ali Mirsoleimani, Jaap van den Herik, Aske Plaat and Jos Vermaseren

Abstract: In this paper, we present a new lock-free tree data structure for parallelMonte Carlo tree search (MCTS) which removes synchronization overhead and guarantees the consistency of computation. It is based on the use of atomic operations and the associated memory ordering guarantees. The proposed parallel algorithm scales very well to higher numbers of cores when compared to the existing methods.

Paper Nr: 115
Title:

Optimizing Super-resolution Reconstruction using a Genetic Algorithm

Authors:

Michal Kawulok, Daniel Kostrzewa, Pawel Benecki and Lukasz Skonieczny

Abstract: Super-resolution reconstruction (SRR) is aimed at increasing spatial resolution given a single image or multiple images presenting the same scene. The existing methods are underpinned with a premise that the observed low resolution images are obtained from a hypothetic high resolution image by applying a certain imaging model (IM) which degrades the image and decreases its resolution. Hence, the reconstruction consists in applying an inverse IM to recover the high resolution data. Such an approach has been found effective, if the IM is known and controlled, in particular when the low resolution images are indeed obtained from a high resolution one. However, in a real-world scenario, when SRR is performed from images originally captured at low resolution, finding appropriate IM and tuning its hyperparameters is a challenging task. In this paper, we propose to optimize the SRR hyperparameters using a genetic algorithm, which has not been reported in the literature so far. We argue that this may substantially improve the capacities of learning the relation between low and high resolution images. Our initial, yet highly encouraging, experimental results reported in the paper allow us to outline our research pathways to deploy the developed techniques in practice.

Paper Nr: 117
Title:

Towards a Query Translation Disambiguation Approach using Possibility Theory

Authors:

Oussama Ben Khiroun, Bilel Elayeb and Narjès Bellamine Ben Saoud

Abstract: We propose in this paper a combined method for Cross-Language Information Retrieval (CLIR) using statistical and lexical resources. On the one hand, we extracted a bilingual French to English dictionary from aligned texts of the Europarl collection. On the other hand, we built a co-occurrence graph structure and used the BabelNet lexical network to process the disambiguation of translation candidates for ambiguous words. We compared our new possibilistic approach with circuit-based one and studied the impact of query expansion by adopting the pseudo-relevance feedback (PRF) technique. Our experiments are performed using the standard CLEF-2003 collection. The results show the positive impact of PRF on the query translation process. Besides, the possibilistic approach using the co-occurrence graph outperforms the overall circuit-based runs.

Paper Nr: 137
Title:

Language Identification of Similar Languages using Recurrent Neural Networks

Authors:

Ermelinda Oro, Massimo Ruffolo and Mostafa Sheikhalishahi

Abstract: The goal of similar Language IDentification (LID) is to quickly and accurately identify the language of the text. It plays an important role in several Natural Language Processing (NLP) applications where it is frequently used as a pre-processing technique. For example, information retrieval systems use LID as a filtering technique to provide users with documents written only in a given language. Although different approaches to this problem have been proposed, similar language identification, in particular applied to short texts, remains a challenging task in NLP. In this paper, a method that combines word vectors representation and Long Short-Term Memory (LSTM) has been implemented. The experimental evaluation on public and well-known datasets has shown that the proposed method improves accuracy and precision of language identification tasks.

Paper Nr: 139
Title:

Usage of Cognitive Architectures in the Development of Industrial Applications - Utilization of a General Cognitive Process in the Domain Building Automation

Authors:

Alexander Wendt, Stefan Kollmann, Lydia Siafara and Yevgen Biletskiy

Abstract: Cognitive architectures, which originate from the field of Artificial Intelligence, implement models for problem-solving and decision-making. These architectures have a wide room for implementation in industrial applications. The goal is to adapt a cognitive architecture to the demands of an application in the area of building automation. It is analyzed, why cognitive architectures are difficult to apply in industrial domain. The result of the analysis is a cognitive process, which is applied to an application in the building automation domain. The use of the architectures is demonstrated within a Java-based based middleware. There, the cognitive architecture is applied for the automatic generation and improvement of control strategies in building automation, which have the goal to minimize energy consumption with minimal reduction of the comfort.

Posters
Paper Nr: 3
Title:

Answering What-type and Who-type Questions for Non-task-oriented Dialogue Agents

Authors:

Makoto Koshinda, Michimasa Inaba and Kenichi Takahashi

Abstract: In this study, we propose a method for responding to what-type and who-type questions with no single fixed response handled by a non-taskoriented dialogue agent using Wikipedia as a language resource. The proposed method extracts nouns from a provided question text and then extracts an article from Wikipedia containing most of those nouns in its title. Next, words are extracted from the extracted article, and the degree of similarity between the extracted words and nouns extracted from question text is calculated. Words with a high degree of similarity are then acquired as response candidates. Next, the response candidates are ranked using the Wikipedia article structure and the text within the article, and the first-place response candidate is used for a response. According to the evaluation experiments, it was confirmed that the proposed method is capable of relatively natural responses in comparison to a baseline.

Paper Nr: 4
Title:

Cardiac Disorder Detection Application and ANT+ Technology

Authors:

Ikram Nedjai Merrouche, Amina Makhlouf, Nadia Saadia and Amar Ramdane-Cherif

Abstract: The problems caused by the occurrence of a heart disorder are great threats to the elderly. With the evolution of new mobile technologies and data transmission, the smartphone has become an ideal platform for the development of applications that can monitoring the person in order to be able to provide assistance if necessary. In order to transmit the real-time data of a cardiac sensor placed on the person to a smartphone, a communication medium is required which consumes preferably the least battery possible. In this article we use a new technology called ANT + that promises a very good rate of wireless transmission with low power consumption. We present a system that offers the doctor or the person in charge of the security of the elderly the possibility of recording different data concerning the person monitored. This data is used in a cardiac disorder detection algorithm, and allow our system to match any type of profile. In addition, we are implementing an Android application, which monitors real-time heartbeat transmitted from a belt using ANT + technology, and detects any heart problems.

Paper Nr: 27
Title:

Exploration Methods for Connectionist Q-learning in Bomberman

Authors:

Joseph Groot Kormelink, Madalina M. Drugan and Marco A. Wiering

Abstract: In this paper, we investigate which exploration method yields the best performance in the game Bomberman. In Bomberman the controlled agent has to kill opponents by placing bombs. The agent is represented by a multi-layer perceptron that learns to play the game with the use of Q-learning. We introduce two novel exploration strategies: Error-Driven-e and Interval-Q, which base their explorative behavior on the temporal-difference error of Q-learning. The learning capabilities of these exploration strategies are compared to five existing methods: Random-Walk, Greedy, e-Greedy, Diminishing e-Greedy, and Max-Boltzmann. The results show that the methods that combine exploration with exploitation perform much better than the Random-Walk and Greedy strategies, which only select exploration or exploitation actions. Furthermore, the results show that Max-Boltzmann exploration performs the best in overall from the different techniques. The Error-Driven-e exploration strategy also performs very well, but suffers from an unstable learning behavior.

Paper Nr: 36
Title:

Disruption Recovery within Agent Organisations in Distributed Systems

Authors:

Asia Al-karkhi and Maria Fasli

Abstract: One of the challenging problems in distributed systems is dealing with agent failure. In this paper, we present an approach for task recovery in a distributed system where the agents self-organise themselves in organisations in order to execute tasks more efficiently. However, within this setting, unpredictable events can happen and agents can fail leading to task failure and weaker overall system performance. We present the Henchman recovery protocol which enables the agents within an organisation to maintain task execution and recover tasks in the event of agent failure. We show how the protocol helps to maintain the efficiency of the created organisations through a series of experiments in a simulated distributed task execution system which has been implemented in Repast Simphony. The experimental results demonstrate the robustness of the proposed solution in a number of settings.

Paper Nr: 37
Title:

Getting More Out of Small Data Sets - Improving the Calibration Performance of Isotonic Regression by Generating More Data

Authors:

Tuomo Alasalmi, Heli Koskimäki, Jaakko Suutala and Juha Röning

Abstract: Often it is necessary to have an accurate estimate of the probability that a classifier prediction is indeed correct. Many classifiers output a prediction score that can be used as an estimate of that probability but for many classifiers these prediction scores are not well calibrated. If enough training data is available, it is possible to post process these scores by learning a mapping from the prediction scores to probabilities. One of the most used calibration algorithms is isotonic regression. This kind of calibration, however, requires a decent amount of training data to not overfit. But many real world data sets do not have excess amount of data that can be set aside for calibration. In this work, we have developed a data generation algorithm to produce more data from a limited sized training data set. We used two variations of this algorithm to generate the calibration data set for isotonic regression calibration and compared the results to the traditional approach of setting aside part of the training data for calibration. Our experimental results suggest that this can be a viable option for smaller data sets if good calibration is essential.

Paper Nr: 41
Title:

Data-driven Relevancy Estimation for Event Logs Exploration and Preprocessing

Authors:

Pierre Dagnely, Elena Tsiporkova and Tom Tourwé

Abstract: With the realization of the industrial IoT, more and more industrial assets are continuously monitored by loggers that report events (states, warnings and failures) occurring in or around these devices. Unfortunately, the amount of events in these event logs prevent an efficient exploration, visualization and advanced exploitation of this data. Therefore, a method that could estimate the relevancy of an event is crucial. In this paper, we propose 10 methods, inspired from various research fields, to estimate event relevancy. These methods have been benchmarked on two industrial datasets composed of event logs from two photovoltaic plants. We have demonstrated that a combination of methods can detect irrelevant events (which can correspond to up to 90% of the data). Hence, this is a promising preprocessing step that can help domain experts to explore the logs in a more efficient way and can optimize the performance of analytical methods by reducing the training dataset size without losing information.

Paper Nr: 42
Title:

Querying Social Practices in Hospital Context

Authors:

John Bruntse Larsen, Virginia Dignum, Jørgen Villadsen and Frank Dignum

Abstract: Understanding the social contexts in which actions and interactions take place is of utmost importance for planning one’s goals and activities. People use social practices as means to make sense of their environment, assessing how that context relates to past, common experiences, culture and capabilities. Social practices can therefore simplify deliberation and planning in complex contexts. In the context of patient-centered planning, hospitals seek means to ensure that patients and their families are at the center of decisions and planning of the healthcare processes. This requires on one hand that patients are aware of the practices being in place at the hospital and on the other hand that hospitals have the means to evaluate and adapt current practices to the needs of the patients. In this paper we apply a framework for formalizing social practices of an organization to an emergency department that carries out patient-centered planning. We indicate how such a formalization can be used to answer operational queries about the expected outcome of operational actions.

Paper Nr: 75
Title:

Ontology-based Information Extraction from Technical Documents

Authors:

Syed Tahseen Raza Rizvi, Dominique Mercier, Stefan Agne, Steffen Erkel, Andreas Dengel and Sheraz Ahmed

Abstract: This paper presents a novel system for extracting user relevant tabular information from documents. The presented system is generic and can be applied to any documents irrespective of their domain and the information they contain. In addition to the generic nature of the presented approach, it is robust and can deal with different document layouts followed while creating those documents. The presented system has two main modules; table detection and ontological information extraction. The table detection module extracts all tables from a given technical document while, the ontological information extraction module extracts only relevant tables from all of the detected tables. The generalization in this system is achieved by using ontologies, thus enabling the system to adapt itself, to a new set of documents from any other domain, according to any provided ontology. Furthermore, the presented system also provides a confidence score and explanation of the score for each of the extracted tables in terms of its relevancy. The system was evaluated on 80 real technical documents of hardware parts containing 2033 tables from 20 different brands of Industrial Boilers domain. The evaluation results show that the presented system extracted all of the relevant tables and achieves an overall precision, recall, and F-measure of 0.88, 1 and 0.93 respectively.

Paper Nr: 77
Title:

SentiMozart: Music Generation based on Emotions

Authors:

Rishi Madhok, Shivali Goel and Shweta Garg

Abstract: Facial expressions are one of the best and the most intuitive way to determine a person’s emotions. They most naturally express how a person is feeling currently. The aim of the proposed framework is to generate music corresponding to the emotion of the person predicted by our model. The proposed framework is divided into two models, the Image Classification Model and the Music Generation Model. The music would be generated by the latter model which is essentially a Doubly Stacked LSTM architecture. This is to be done after classification and identification of the facial expression into one of the seven major sentiment categories: Angry, Disgust, Fear, Happy, Sad, Surprise and Neutral, which would be done by using Convolutional Neural Networks (CNN). Finally, we evaluate the performance of our proposed framework using the emotional Mean Opinion Score (MOS) which is a popular evaluation metric for audio-visual data.

Paper Nr: 89
Title:

Polish Texts Topic Classification Evaluation

Authors:

Tomasz Walkowiak and Piotr Malak

Abstract: Abstract: The paper presents preparation, lead and results of evaluation of efficiency of text classification (TC) methods for Polish. The subject language is of complex morphology, it belongs to flexional languages. Thus there is a strong need of making proper text preprocessing in order to guarantee reliable TC. Basing on authors’ practical experience from former TC, IR and general NLP experiments set of preprocessing rules was applied. Also feature-documents matrix was designed with respect to the most promising feature selected. About 216 experiments on exemplar corpus in subject (topic) classification task, with different preprocessing, weighting, filtering (for dimensions reduction) schemes and classifiers was conducted. Results shows there is not substantial increase of accuracy when using most of classical pre-processing steps in case of corpus of large size (at least 1000 exemplars per class). The highest impact authors were able to obtain concerned the system costs of TC processes, not the TC accuracy.

Paper Nr: 97
Title:

Evolutionary Clustering Techniques for Expertise Mining Scenarios

Authors:

Veselka Boeva, Milena Angelova, Niklas Lavesson, Oliver Rosander and Elena Tsiporkova

Abstract: The problem addressed in this article concerns the development of evolutionary clustering techniques that can be applied to adapt the existing clustering solution to a clustering of newly collected data elements. We are interested in clustering approaches that are specially suited for adapting clustering solutions in the expertise retrieval domain. This interest is inspired by practical applications such as expertise retrieval systems where the information available in the system database is periodically updated by extracting new data. The experts available in the system database are usually partitioned into a number of disjoint subject categories. It is becoming impractical to re-cluster this large volume of available information. Therefore, the objective is to update the existing expert partitioning by the clustering produced on the newly extracted experts. Three different evolutionary clustering techniques are considered to be suitable for this scenario. The proposed techniques are initially evaluated by applying the algorithms on data extracted from the PubMed repository.

Paper Nr: 108
Title:

Cylindric Clock Model to Represent Spatio-temporal Trajectories

Authors:

Joanna Isabelle Olszewska

Abstract: To automatically understand agents’ environment and its changes, the study of spatio-temporal relations between the objects evolving in the observed scene is of prime importance. In particular, the temporal aspect is crucial to analyze scene’s objects of interest and their trajectories, e.g. to follow their movements, understand their behaviours, etc. In this paper, we propose to conceptualize qualitative spatio-temporal relations in terms of the clock model and extend it to a new spatio-temporal model we called cylindric clock model, in order to effectively perform automated reasoning about the scene and its objects of interest and to improve the modeling of dynamic scenes compared to state-of-art approaches as demonstrated in the carried out experiments. Hence, the new formalisation of the qualitative spatio-temporal relations provides an efficient method for both knowledge representation and information processing of spatio-temporal motion data.

Paper Nr: 111
Title:

Towards a Stable Quantized Convolutional Neural Networks: An Embedded Perspective

Authors:

Motaz Al-Hami, Marcin Pietron, Raul Casas, Samer Hijazi and Piyush Kaul

Abstract: Nowadays, convolutional neural network (CNN) plays a major role in the embedded computing environment. Ability to enhance the CNN implementation and performance for embedded devices is an urgent demand. Compressing the network layers parameters and outputs into a suitable precision formats would reduce the required storage and computation cycles in embedded devices. Such enhancement can drastically reduce the consumed power and the required resources, and ultimately reduces cost. In this article, we propose several quantization techniques for quantizing several CNN networks. With a minor degradation of the floating-point performance, the presented quantization methods are able to produce a stable performance fixed-point networks. A precise fixed point calculation for coefficients, input/output signals and accumulators are considered in the quantization process.

Paper Nr: 112
Title:

Modelling Attitudes of Dialogue Participants - Reasoning and Communicative Space

Authors:

Mare Koit

Abstract: The paper introduces a dialogue model, concentrating on attitudes of dialogue participants. Two kinds of attitudes are under consideration: (1) attitudes related to different aspects of a negotiation object (in our case, doing an action) which direct reasoning in communication, and (2) attitudes related to a communication partner (dominance-subordination, cooperation-antagonism, communicative distance, etc.) which are modelled by using the concept of communicative space. Telemarketing calls in the Estonian dialogue corpus are analysed in order to illustrate communicative space and to find out linguistic cues for automatic recognition of different coordinates. A limited version of the dialogue model is implemented on the computer.

Paper Nr: 119
Title:

Pipeline Pattern for Parallel MCTS

Authors:

S. Ali Mirsoleimani, Jaap van den Herik, Aske Plaat and Jos Vermaseren

Abstract: In this paper, we present a new algorithm for parallel Monte Carlo tree search (MCTS). It is based on the pipeline pattern and allows flexible management of the control flow of the operations in parallel MCTS. The pipeline pattern provides for the first structured parallel programming approach to MCTS. The Pipeline Pattern for Parallel MCTS algorithm (called 3PMCTS) scales very well to a higher number of cores when compared to the existing methods. The observed speedup is 21 on a 24-core machine.

Paper Nr: 126
Title:

Ontology Design for Task Allocation and Management in Urban Search and Rescue Missions

Authors:

Elie Saad, Koen V. Hindriks and Mark A. Neerincx

Abstract: Task allocation and management is crucial for human-robot collaboration in Urban Search And Rescue response efforts. The job of a mission team leader in managing tasks becomes complicated when adding multiple and different types of robots to the team. Therefore, to effectively accomplish mission objectives, shared situation awareness and task management support are essential. In this paper, we design and evaluate an ontology which provides a common vocabulary between team members, both humans and robots. The ontology is used for facilitating data sharing and mission execution, and providing the required automated task management support. Relevant domain entities, tasks, and their relationships are modeled in an ontology based on vocabulary commonly used by firemen, and a user interface is designed to provide task tracking and monitoring. The ontology design and interface are deployed in a search and rescue system and its use is evaluated by firemen in a task allocation and management scenario. Results provide support that the proposed ontology (1) facilitates information sharing during missions; (2) assists the team leader in task allocation and management; and (3) provides automated support for managing an Urban Search and Rescue mission.

Paper Nr: 129
Title:

Error Correction for Information Retrieval of Czech Documents

Authors:

Jiří Martínek and Pavel Král

Abstract: This paper proposes a novel system for information retrieval over a set of scanned documents in the Czech language. The documents are in the form of raster images and thus they are first converted into the text form by optical character recognition (OCR). Then OCR errors are corrected and the corrected texts are indexed and stored into a fulltext database. The database provides a possibility of searching over these documents. This paper describes all components of the above mentioned system with a particular focus on the proposed OCR correction method. We experimentally show that the proposed approach is efficient, because it corrects a significant number of errors. We also create a small Czech corpus to evaluate OCR error correction methods which represent another contribution of this paper.

Area 2 - Agents

Full Papers
Paper Nr: 18
Title:

Detecting Influence in Wisdom of the Crowds

Authors:

Luís Correia, Sofia Silva and Ana Cristina B. Garcia

Abstract: The wisdom of the crowds effect (WoC) is a collective intelligence (CI) property by which, given a problem, a crowd is able to provide a solution better than that of any of its individuals. However, WoC is considered to require that participants are not priorly influenced by information received on the subject of the problem. Therefore it is important to have metrics that can identify the presence of influence in an experiment, so that who runs it can decide if the outcome is product of the WoC or of a cascade of individuals influencing others. In this paper we provide a set of metrics that can analyse a WoC experiment as a data stream and produce a clear indication of the presence of some influence. The results presented were obtained with real data from different information conditions, and are encouraging. The paper concludes with a discussion of relevant situations and points the most important steps that follow in this research.

Paper Nr: 51
Title:

Team Distribution between Foraging Tasks with Environmental Aids to Increase Autonomy

Authors:

Juan M. Nogales and Gina Maira Barbosa de Oliveira

Abstract: In this paper, robots have to distribute themselves across a set of regions where they will serve in foraging tasks, transporting objects repetitively. Each region stores information about the performance of the subgroup of robots serving that region. Robots can also share information between them and identify which region is offering better conditions to forage. In particular, each region has a different rate to recover recently removed objects, which demands a different number of robot foragers. We explore the effects of the network structure in robot distribution and their performance. Results indicate a small dependence of robot-robot connections and a great dependence of robot-environment interaction. Since cooperative robots are going after a global goal, the proposed distribution rules combined with environmental aids allowed them to make better decisions autonomously, increasing the number of transported objects and reducing the number of travels.

Paper Nr: 52
Title:

Study of Route Optimization Considering Bottlenecks and Fairness Among Partial Paths

Authors:

Toshihiro Matsui, Marius Silaghi, Katsutoshi Hirayama, Makoto Yokoo and Hiroshi Matsuo

Abstract: Route optimization is an important problem for single agents and multi-agent systems. In route optimization tasks, the considered challenges generally belong to the family of shortest path problems. Such problems are solved using optimization algorithms, such as the A* algorithm, which is based on tree search and dynamic programming. In several practical cases, cost values should be as evenly minimized for individual parts of paths as possible. These situations are also considered as multi-objective problems for partial paths. Since dynamic programming approaches are employed for the shortest path problems, different types of criteria which can be decomposed with dynamic programming might be applied to the conventional solution methods. For this class of problems, we employ a leximax-based criterion, which considers the bottlenecks and unfairness among the cost values of partial paths. This criterion is based on a similar criterion called leximin for multi-objective maximization problems. It is also generalized for objective vectors which have variable lengths. We address an extension of the conventional A* search algorithm and investigate an issue concerning on-line search algorithms. The influence of the proposed approach is experimentally evaluated.

Paper Nr: 58
Title:

CARS - A Spatio-temporal BDI Recommender System: Time, Space and Uncertainty

Authors:

Amel Ben Othmane, Andrea Tettamanzi, Serena Villata and Nhan Le Thanh

Abstract: Agent-based recommender systems have been exploited in the last years to provide informative suggestions to users, showing the advantage of exploiting components like beliefs, goals and trust in the recommendations’ computation. However, many real-world scenarios, like the traffic one, require the additional feature of representing and reasoning about spatial and temporal knowledge, considering also their vague connotation. This paper tackles this challenge and introduces CARS, a spatio-temporal agent-based recommender system based on the Belief-Desire-Intention (BDI) architecture. Our approach extends the BDI model with spatial and temporal information to represent and reason about fuzzy beliefs and desires dynamics. An experimental evaluation about spatio-temporal reasoning in the traffic domain is carried out using the NetLogo platform, showing the improvements our recommender system introduces to support agents in achieving their goals.

Paper Nr: 67
Title:

Approximate Algorithms for Double Combinatorial Auctions for Resource Allocation in Clouds: An Empirical Comparison

Authors:

Diana Gudu, Gabriel Zachmann, Marcus Hardt and Achim Streit

Abstract: There is an increasing trend towards market-driven resource allocation in cloud computing, which can address customer requirements for flexibility, fine-grained allocation, as well as improve provider revenues. We formulate the cloud resource allocation as a double combinatorial auction. However, combinatorial auctions are NP-hard problems. Determining the allocation optimally is thus intractable in most cases. Various heuristics have been proposed, but their performance and quality of the obtained solutions are highly dependent on the input. In this paper, we perform an extensive empirical comparison of several approximate allocation algorithms for double combinatorial auctions. We discuss their performance, economic efficiency, and the reasons behind the observed variations in approximation quality. Finally, we show that there is no clear winner: no algorithm outperforms the others in all test scenarios. Furthermore, we introduce a novel artificial input generator for combinatorial auctions which uses parameterized random distributions for bundle sizes, resource type selection inside a bundle, and the bid values and reserve prices. We showcase its flexibility, required for thorough benchmark design, through a wide range of test cases.

Paper Nr: 80
Title:

A Novel Tool for Detecting Indirect Normative Conflicts in Multi-agent Systems

Authors:

Jéssica Soares dos Santos and Viviane Torres da Silva

Abstract: Norms are usually applied in Multi-Agent Systems to regulate the behavior of software agents and maintain social order. Those systems can be regulated by multiple norms and require a mechanism to verify whether the set of norms is conflict-free or not. The detection of indirect normative conflicts is not a trivial task since they only can be identified when the detection mechanism is able to infer that different elements that compose two norms are related in some way. In this research, we propose a mechanism to detect normative conflicts by combining two different approaches. The former uses information from a domain ontology that stores relationships that are exclusive of the MAS. The latter uses information from a lexical database called WordNet that stores relationships among concepts of the real world. This research results in the implementation of a tool with a robust mechanism for normative conflict detection that can be used during the design of a MAS.

Paper Nr: 83
Title:

An Architecture for Autonomous Normative BDI Agents based on Personality Traits to Solve Normative Conflicts

Authors:

Paulo Henrique Cardoso Alves, Marx Leles Viana and Carlos José Pereira de Lucena

Abstract: Norms are promising mechanisms of social control to ensure a desirable social order in open multiagent systems. Normative multiagent systems offer the ability to integrate social and individual factors to provide increased levels of fidelity with respect to modelling social phenomena such as cooperation; coordination; decision-making process, and organization in artificial agent systems. However, norms eventually can be conflicting — for example, when there is a norm that prohibits an agent to perform a particular action and another norm that obligates the same agent to perform the same action, the agent is not able to fulfill both norms at the same time. The agent’s decision about which norms to fulfill can be defined based on rewards, punishments and agent’s goals. Sometimes, the analysis between these attributes will not be enough to allow the agent to make the best decision. This paper introduces an architecture that considers the agent’s personality traits in order to improve the normative conflict solving process. In addition, the agent can execute different behaviors with equal environment variables, just by changing its own internal characteristics. The applicability and validation of our approach are demonstrated by an experiment that reinforces the importance of the society’s norms.

Paper Nr: 84
Title:

RPI.Social: Simple Enactment and Execution of First-Class Agent Interaction Protocols

Authors:

Atef Nouri and Wided Lejouad Chaari

Abstract: In Multiagent systems, first-class interaction protocols are those whose implementations are decoupled from the agents. A previous work has introduced the RPI framework (Role Playing Interactions) and established the contribution of RPI.Idiom which is a high-level language for the definition of such protocols. In this paper, we present RPI.Social which is a social engine associated with the agent to help it use interaction protocols written in RPI.Idiom. RPI.Social has two components: RPI.Social.IoP (for Initiate or Play) and RPI.Social.Exec. The first component deals with discovering and initiating interaction protocols for an agent which has a goal that could only be pursued through interacting. The same component is also used by agents invited for interaction to decide whether to participate or to decline the invitation. The second component serves as an interpreter of interaction protocols with several mechanisms and rules to coordinate and share results amongst the interacting agents. The main contribution of this paper is a solution for agents to automatically identify and execute first-class interaction protocols.

Paper Nr: 109
Title:

Multi-Agent-Based Generation of Explanations for Retrieval Results Within a Case-Based Support Framework for Architectural Design

Authors:

Viktor Ayzenshtadt, Christian Espinoza-Stapelfeld, Christoph Langenhan and Klaus-Dieter Althoff

Abstract: In this paper, we describe the general structure and evaluation of a multi-agent based system module that was conceptualized to explain, and therefore, enrich the search results of the retrieval process within a distributed case-based framework for support of early conceptual design phase in architecture. This explanation module is implemented as an essential part of the framework and uses case-based agents, explanation ontology, and explanation patterns as its underlying foundational components. The module’s main goal is to provide the user with additional information about the search results to make the framework’s behavior during the retrieval stage more transparent and traceable. System’s justification for displaying of results plays an important role as well, and is also included in the explanations. We evaluated the explanation generation process with a ground-truth set of explanations and a case-based validation process to ensure the suitability of the generated explanation expressions for displaying in user interfaces connected to the framework. The results of the evaluation confirmed our expectation and showed the general validity of the explanations.

Paper Nr: 124
Title:

A Stochastic Model of Diffusion in Opinion Dynamics

Authors:

Stefania Monica and Federico Bergenti

Abstract: This paper studies analytically the dynamics of the opinion in multi-agent systems when only the sociological phenomenon known as diffusion is considered. First, the paper recalls a framework for the analytic study of opinion dynamics which has been already applied to describe the effects of a number of sociological phenomena. Then, the framework is specialized to the study of diffusion, according to which the opinion of an agent can be influenced by the social context. Diffusion is introduced in the framework by stating stochastic rules meant to describe at the microscopic level how diffusion contributes to change the opinion of an agent. The obtained model is used to derive collective and asymptotic properties of multi-agent systems when only diffusion is considered, which are verified against specific simulations shown in the last part of the paper. The paper is concluded with a recapitulation of presented results and an outline of future work.

Paper Nr: 149
Title:

Integration Between Agents and Remote Ontologies for the Use of Content on the Semantic Web

Authors:

Felipe Demarchi, Elder Rizzon Santos and Ricardo Azambuja Silveira

Abstract: The Semantic Web proposes a structure of significant content for Web pages that is used in knowledge bases and developed from ontologies, that have recently come to coexist on the Web. There are studies to allow agents to navigate through these knowledge bases in search of answers to queries. This work proposes the adaptation of a well-known agent structure, named Jason, in order to allow the agent access to ontologies available on the Web. In this context, efforts have been made to perform the integration of agents with ontologies, most of which allow the knowledge of the agent to be based on a local ontology. However, applying the ability to use semantic data available on the Web to a consolidated belief-desire-intention (BDI) agent structure is a subject that still needs to be explored. Therefore, this work proposes changes in the implementation of the Jason interpreter that would allow agents to access ontologies available on the Web to perform the update of their belief base based on significant content. As validation, a case study of an educational quiz is presented that uses this information to formulate the questions and validate the answers obtained.

Paper Nr: 155
Title:

Agent-based Simulation of Socially-inspired Model of Resistance against Unpopular Norms

Authors:

Arshad Muhammad, Kashif Zia and Dinesh Kumar Saini

Abstract: People lives in the society adhering to different norms. Some of these norms are unpopular. Sometimes, for the overall societal good, it is necessary to oppose and possibly avert unpopular norms. To achieve this goal, it is necessary to know the conditions, which enable persistence of the unpopular norms and models that support possible aversion of them. This study attempts to elaborate the conditions and reasons for the emergence, spreading and aversion of unpopular norms in society, using theory-driven agent-based simulation. The simulation results reveal that in addition to agents actively participating in averting the unpopular norm, incorporating a rational decision-making model in the population of agents is necessary to achieve a dominant norm aversion. The significance of these results concerning digital societies is enormous. In the new social landscape dominated by digital contents (particularly of social networking), it can be argued that careful amalgamation of social media contents can not only educate the people but also be useful in aversion of undesirable behavior, for example, retention and spreading of unpopular norms.

Short Papers
Paper Nr: 34
Title:

Agent-based Simulation Model Embedded Accounting’s Purchase Method; Analysis on the Systemic Risk of Mergers and Acquisitions between Financial Institutions

Authors:

Hidenori Kato, Yuichi Sei, Yasuyuki Tahara and Akihiko Ohsuga

Abstract: The aim of the present study is to evaluate systemic risks due to merger between financial institutions by manipulating the decline rate of marketable asset price. An agent-based simulation platform with purchase method of international financial reporting standards (IFRS) is developed and analyses the influence of the goodwill (Noren), produced by mergers between financial institutions. The research reveals the following two points: (1) the decline rate of marketable asset price determines the number of bankruptcies, (2) when market value asset price plumps sharply, the effect of merger is small.

Paper Nr: 48
Title:

Using Blockchains for Agent-based Auctions

Authors:

Leo van Moergestel, Mathijs van Bremen, Bianca Krieger, Martijn van Dijk and Erik Puik

Abstract: To extend the lifetime of products, an agent is connected to the product. This agent has several roles. It depends on the phase of the lifecycle what these roles will be. One of the roles in the usage or recycling phase is to negotiate for buying spare parts in case a part of the product is broken. The same agent can also decide to offer spare parts to other agents to reuse working parts of a broken product. To accomplish this idea, a marketplace for agents has to be set up, where the auctions can take place. To support this concept, blockchain technology has been used. Blockchains are a new type of technology, known from bitcoins, but there are other cases where blockchains can be used. Blockchain is known for its decentralisation, transparency and for making trustful transactions. In this paper the working of different types of blockchains will be briefly explained and determined if they can be useful for online auctions by agents. A prototype of the marketplace using blockchains has been built.

Paper Nr: 62
Title:

Role of Trust in Creating Opinions in Social Networks

Authors:

Jiří Jelínek

Abstract: Although we are not always aware of this, our existence and especially communication are based on the principles of trust. The importance of trust is crucial in systems where risk is present e.g. when handling the information we have acquired through communication because it is not always possible to immediately verify the truthfulness of it. The aim of this paper is to link two areas, namely the reality in the human community described above and the available knowledge of social networks and multi-agent systems, and try to simulate real trust concerned scenarios in society by these tools. The multi-agent model will be presented, which simulates the behavior of the heterogeneous group (people-like) entities in the process of creating their opinion about the world on the basis of information acquired through communication with other agents. The focus is placed on the processes influencing trust in communication partners and its dynamics. The results of the experiments are also presented.

Paper Nr: 65
Title:

Strategies for Resolving Normative Conflict That Depends on Execution Order of Runtime Events in Multi-Agent Systems

Authors:

Mairon Belchior, Jéssica Soares dos Santos and Viviane Torres da Silva

Abstract: Norms are being used in multi-agent systems to control the behavior of software agents and maintain social order. They define which actions each agent can or not perform in different circumstances. Systems regulated by multiple norms must be able to detect and resolve normative conflicts to guarantee the expected behavior of the system. A normative conflict arises when a given agent is prohibited and obliged to perform the same action at the same time. Our work aims to resolve normative conflict that occurs at runtime and where its detection depends on the execution order of runtime events in multi-agent systems. This paper presents two independent approaches to resolve the conflicts. The first approach resolves the conflict at design time by eliminating the overlaps between two norms in conflict. The second approach resolves the normative conflict at runtime by extending an existing automated planning algorithm in order to get plans that do not produce sequence of conflicting actions.

Paper Nr: 68
Title:

Multi-Agent Systems' Negotiation Protocols for Cyber-Physical Systems: Results from a Systematic Literature Review

Authors:

Davide Calvaresi, Kevin Appoggetti, Luca Lustrissimi, Mauro Marinoni, Paolo Sernani, Aldo F. Dragoni and Michael Schumacher

Abstract: Cyber Physical Systems (CPS) require a multitude of components interacting among themselves and with the users to perform automatic actions, usually under unpredictable or uncertain conditions. Multi-Agent Systems (MAS) have emerged over the years as one of the major technological paradigms regulating interactions and negotiations among autonomous entities running under heterogeneous conditions. As such, MAS have the potential to support CPS in implementing a highly reconfigurable distributed thinking. However, some gaps are still present between MAS’ features and the strict requirements of CPS. The most relevant is the lack of reliability, which is mainly due to specific features characterizing negotiation protocols. This paper presents a systematic literature review of MAS negotiation protocols aiming at providing a comprehensive overview of their strengths and limitations, examining both the assumptions and requirements set during their development. While this work confirms the potential of MAS in regulating the interactions among CPS components, the findings also highlight the absence of real-time compliance in current negotiation protocols. Strongly characterizing CPS, the capability to face strict time constraints could bridge the gap between MAS and CPS.

Paper Nr: 76
Title:

Modeling Personality in the Affective Agent Architecture GenIA3

Authors:

Joaquin Taverner, Bexy Alfonso, Emilio Vivancos and Vicente Botti

Abstract: In the last few years there has been a growing interest in affective computing. This type of computation tries to include and use emotions in different software processes. One of the most relevant areas is the simulation of human behavior where various affective models are used to represent different affective characteristics such as emotions, mood, or personality. Personality is defined as a set of individual characteristics that influence motivations and behaviors when a human being faces a particular circumstance. Personality plays a very important role in modeling affective processes. Through the simulation of emotions we can improve, among others, the experience of users dealing with machines, and human simulations in decision-making processes using multi-agent systems. In this work we propose a model for the use of personality in the general purpose architecture for affective agents GenIA3, as well as the development of the model in the current GenIA3 platform.

Paper Nr: 87
Title:

Domain-specific Trust for Context-aware BDI Agents - Preliminary Work

Authors:

Arthur Casals, Eduardo Fermé and Anarosa A. F. Brandão

Abstract: Context-aware systems are capable of perceiving the physical environment where they are deployed and adapt their behavior accordingly. Multiagent systems based on the BDI architecture can be used to process contextual information in the form of beliefs. Contextual information can be divided and structured in the form of information domains. Information and experience sharing enables a single agent to receive data on different information domains from another agent. In this scenario, establishing a trust model between agents can take into account the relative perceptions each agent has of the others, as well as different trust degrees for different information domains. The objective of this work is to adapt an epistemic model to be used by agents with their belief revision in order to establish a mechanism of domain-specific relative trust attribution. Such mechanism will allow for each agent to possess different trust degrees associated with other agents regarding different information domains.

Paper Nr: 106
Title:

Comparison Between Static and Dynamic Willingness to Interact in Adaptive Autonomous Agents

Authors:

Mirgita Frasheri, Baran Cürüklü and Mikael Ekström

Abstract: Adaptive autonomy (AA) is a behavior that allows agents to change their autonomy levels by reasoning on their circumstances. Previous work has modeled AA through the willingness to interact, composed of willingness to ask and give assistance. The aim of this paper is to investigate, through computer simulations, the behavior of agents given the proposed computational model with respect to different initial configurations, and level of dependencies between agents. Dependency refers to the need for help that one agent has. Such need can be fulfilled by deciding to depend on other agents. Results show that, firstly, agents whose willingness to interact changes during run-time perform better compared to those with static willingness parameters, i.e. willingness with fixed values. Secondly, two strategies for updating the willingness are compared, (i) the same fixed value is updated on each interaction, (ii) update is done on the previous calculated value. The maximum number of completed tasks which need assistance is achieved for (i), given specific initial configurations.

Paper Nr: 107
Title:

Evaluation of Dishonest Argumentation based on an Opponent Model: A Preliminary Report

Authors:

Kazuyuki Kokusho and Kazuko Takahashi

Abstract: This paper discusses persuasive dialogue in a case where dishonesty is permitted. We have previously proposed a dialogue model based on a predicted opponent model using an abstract argumentation framework, and discussed the conditions under which a dishonest argument could be accepted without being detected. However, it is hard to estimate the outcome of a dialogue, or identify causality between agents’ knowledge and the result. In this paper, we implement our dialogue model and execute argumentations between agents under different conditions. We analyze the results of these experiments and discuss about them. In brief, our results show that the use of dishonest arguments affects the likelihood of successfully persuading the opponent, or winning a debate game, but we could not identify a relationship between the results of a dialogue and the initial argumentation frameworks of the agents.

Paper Nr: 128
Title:

A Normative Agent-based Model for Sharing Data in Secure Trustworthy Digital Market Places

Authors:

Ameneh Deljoo, Tom van Engers, Robert van Doesburg, Leon Gommans and Cees de Laat

Abstract: Norms are driving forces in social systems and governing many aspects of individual and group decision-making. Various scholars use agent based models for modeling such social systems, however, the normative component of these models is often neglected or relies on oversimplified probabilistic models. Within the multi-agent research community, the study of norm emergence, compliance and adoption has resulted in new architectures and standards for normative agents. We propose the N-BDI* architecture by extending the Belief-Desire and Intention (BDI) agents’ control loop, for constructing normative agents to model social systems; the aim of our research to create a better basis for studying the effects of norms on a society of agents. In this paper, we focus on how norms can be used to create so-called Secure Trustworthy Digital Marketplaces (STDMPs). We also present a case study showing the usage of our architecture for monitoring the STDMP-members’ behavior. As a concrete result, a preliminary implementation of the STDMP framework has been implemented in multi-agent systems based on Jadex.

Paper Nr: 150
Title:

Effective Evaluation of Autonomous Taxi Fleets

Authors:

Philippe Mathieu and Antoine Nongaillard

Abstract: With the advent of autonomous vehicles, self-management of taxis fleet becomes an important issue for the automotive industry. Designing strategies for taxis turns out to be a difficult task due to a large number of parameters and metrics involved. Performance evaluation of these strategies is also a complex problem since effectiveness in some configurations may become inefficiency in others. After formalizing this problem we propose several strategies based on swarm-computing techniques. Finally, we show that metric unification is necessary and that only a multi-criterion approach illustrated by an economic analysis allows a comparison. We conclude with a description of the simulator implemented and some examples showing the measurements made with the proposed strategies.

Posters
Paper Nr: 12
Title:

Speeding up the Search of a Global Dynamic Equilibrium from a Local Cooperative Decision

Authors:

Sébastien Maignan, Carole Bernon and Pierre Glize

Abstract: Systems composed of many interdependent active entities working on shared resources can be challenging to regulate and multi-agent simulation is an efficient means for finding the suitable entities’ behaviors. On the other hand, the search space for a stable solution of the system is usually forbidding and hinders any effort to solve the problem using a top-down approach. Furthermore, the complexity of possible global functions to optimize increases rapidly with the size of the system and can prove difficult to define and/or evaluate at each system simulation step. The difficulty when designing bottom-up systems is to be able to identify all their emergent properties and the parameters to modulate them. Here we propose a local cooperative decision making process that helps to stabilize such systems. These local processes prove to be very efficient to quickly find dynamic equilibrium solutions where the system continues to function and fulfills its global function. Regulation emerges from simple local interactions.

Paper Nr: 15
Title:

Searching for Effective and Efficient Way of Knowledge Transfer within an Organization

Authors:

Agnieszka Kowalska-Styczeń, Krzysztof Malarz and Kamil Paradowski

Abstract: In this paper three models of knowledge transfer in organization are considered. In the first model (A) the transfer of chunks of knowledge among agents is possible only when the sender has exactly one more chunks of knowledge than recipient. This is not dissimilar with bounded confidence model of opinion dynamics. In the second model (B) the knowledge transfer take place when sender is “smarter” than recipient. Finally, in the third scenario (model C) we allow for knowledge transfer also when sender posses the same or greater number of chunks of knowledge as recipient. The simulation bases on cellular automata technique. The organization members occupy nodes of square lattice and they interact only with their nearest neighbors. With computer simulations we show, that the efficiency and the effectiveness of knowledge transfer i) for model C is better than for model B ii) and it is worse for model A than for model B.

Paper Nr: 23
Title:

A MAS Model Approach to a Wind Farm Maintenance Strategy

Authors:

Miguel Kpakpo, Mhamed Itmi and Alain Cardon

Abstract: The aim of this work is to propose a new method of analysis and optimization of maintenance strategy for wind farms. The objective is to help wind farm operator to carry out the optimization of the maintenance costs through profitability analysis of the wind farm according to failures, planned shutdown situations and maintenance budgets. Such approach has the advantage of combining the O&M (optimization and maintenance) technical vision and the financial vision within the meaning of profitability. The platform model is based on multi-agent systems. It aims to realize the calculation and optimization of scenarios. Agents have been identified from the knowledge of the windfarm O&M domain thanks to the wind farm operator’s point of view. The platform we’re developing is named PROMEEO, a French acronym for O&M onshore wind farms rationalization and optimization’s platform.

Paper Nr: 43
Title:

An Agent-based Electronic Market to Help Airlines to Recover from Delays

Authors:

Luís Reis, Ana Paula Rocha and Antonio J. M. Castro

Abstract: The Airline Operations Control Center (AOCC) has the responsibility to ensure that flights meet their planned schedule or, if any problem arises, to find a viable solution that minimizes both the impact in the operational plan and its cost. The high cost of resources involved in this process (aircraft and crew members) leads to a lack of additional resources from the airline companies, implying a restricted solution space. Here, we propose an electronic market modeled as a multi-agent system where airline companies can negotiate and lease each other the required resources when solving a disruption problem, thus expanding their solution space. The proposed negotiation occurs in several rounds, where qualitative comments made by the buyer agent on proposals sent by the sellers enables these to learn how to calculate new proposals, using a case-based reasoning methodology.

Paper Nr: 45
Title:

Area Protection in Adversarial Path-finding Scenarios with Multiple Mobile Agents on Graphs - A Theoretical and Experimental Study of Strategies for Defense Coordination

Authors:

Marika Ivanová, Pavel Surynek and Katsutoshi Hirayama

Abstract: We address a problem of area protection in graph-based scenarios with multiple agents. The problem consists of two adversarial teams of agents that move in an undirected graph. Agents are placed in vertices of the graph and they can move into adjacent vertices in a conflict-free way in an indented environment. The aim of one team - attackers - is to invade into a given area while the aim of the opponent team - defenders - is to protect the area from being entered by attackers. We study strategies for assigning vertices to be occupied by the team of defenders in order to block attacking agents. We show that the decision version of the problem of area protection is PSPACE-hard. Further, we develop various on-line vertex-allocation strategies for the defender team and evaluate their performance in multiple benchmarks. Our most advanced method tries to capture bottlenecks in the graph that are frequently used by the attackers during their movement. The performed experimental evaluation suggests that this method often defends the area successfully even in instances where the attackers significantly outnumber the defenders.

Paper Nr: 56
Title:

Intelligent and Distributed Solving of Multiphysics Problems Coordinated by Software Agents - An Intelligent Approach for Decentralized Simulations

Authors:

Desirée Vögeli, Sebastian Grabmaier, Matthias Jüttner, Michael Weyrich, Peter Göhner and Wolfgang M. Rucker

Abstract: This paper presents an intelligent approach to support engineers with performing computational simulation of new developments and prototypes. With multiple interacting physical effects and large three dimensional models the choice of the right solution strategy is crucial for a correct solution and an acceptable calculation time. The presented multi-agent system can solve these simulation tasks using distributed heterogeneous computation resources with the objective to reduce the calculation time. An important factor for the criterion time is the choice of the linear solver. Here a case-based reasoning concept is introduced to improve the decisions in the multi-agent system. Allowing each agent to solve its problem part by using appropriate solution methods, a decentralized architecture with autonomous software agents is provided.

Paper Nr: 98
Title:

Designing Distributed Multi-Agent System for Aggregate and Final Assembly of Complex Technical Objects on Ramp-up Stage

Authors:

Petr Skobelev, Valery Eliseev, Igor Mayorov, Vitaly Travin, Alexey Zhilyaev and Elena Simonova

Abstract: The paper covers the problem of aggregate and final assembly of complex technical objects at the ramp-up stage. New models, methods and tools for distributed scheduling are proposed, including modified version of virtual market with new classes of agents. The new feature of multi-agent scheduler considers knowledge base technology which helps to specify each operation in more detailed and individual way. The paper describes first system prototype for adaptive planning at the ramp-up stage and the main directions of future system development.

Paper Nr: 116
Title:

Power Optimization by Cooling Photovoltaic Plants as a Dynamic Self-adaptive Regulation Problem

Authors:

Valerian Guivarch, Carole Bernon and Marie-Pierre Gleizes

Abstract: This paper shows an approach to control cooling devices for photovoltaic plants in order to optimize the energy production thanks to a limited reserve of harvested rainwater. This is a complex problem, considering the dynamic environment and the interdependence of the parameters, such as the weather data and the state of the photovoltaic panels. Our claim is to design a system composed of autonomous components cooperating in order to obtain an emergent efficient control.

Paper Nr: 122
Title:

An Illustrative Example of the JADEL Programming Language

Authors:

Eleonora Iotti, Federico Bergenti and Agostino Poggi

Abstract: This paper presents a case study intended to investigate the features of JADEL, an agent-oriented programming language designed to ease the development of JADE agents and multi-agent systems. The paper first motivates the need for JADEL, and it briefly shows the main features of the language. Then, a well-known problem originally designed to assess the features of actor-based programming languages is recalled, and a possible solution implemented in JADEL is presented. The proposed solution is intended to validate the features of the language that concern concurrency and distribution, and it can be used as a guideline to use JADEL to target problems expressed in terms of agents that cooperate to bring about joint goals.