Search

In this paper we investigate the board game Cram, which isan impartial combinatorial game, using an alpha-beta solver. Since Cram is anotoriously hard game in the sense that it is difficult to obtain reliableand useful domain knowledge to use in the search process, we decided torely on knowledge obtained from Combinatorial Game Theory (CGT).The first and most effective addition to our solver is to incorporateendgame databases prefilled with CGT values (nimbers) for all positionsfitting on boards with at most 30 squares. This together with twoefficient move-ordering heuristics aiming at early splitting positions intofragments fitting in the available databases gives a large improvement ofsolving power. Next we define five more heuristics based on CGT thatcan be used to further reduce the sizes of the search trees considerably.In the final version of our program we were able to solve all odd x oddCram boards for which results were available from the literature (evenx even and odd x even boards are trivially solved). Investigating newboards led to solving two boards not solved before, namely the 3 x 21board, a first-player win, and the 5 x 11 board, a second-player win.

Agent-oriented solutions form a useful paradigm to design intelligent systems. For health-related behaviour change, this is also a promising approach. Designing an agent for lifestyle change interventions is a difficult task because socio-ecological models are involved that represent many conflicting desires and goals. Different types of cognitive architectures are available to design this type of health behavior agents but they are rarely used. In this paper, we used the BDI model to design a health behavior agent that will execute behavior change intervention for a better healthy lifestyle. We explain the working of the architecture by the example of an agent which uses adaptive goals-setting and a percentile scheduling technique for increasing physical activity.

In this paper, scientific species names from images of handwritten species observations are automatically recognised and annotated with semantic concepts, so that they can be used for document retrieval and faceted search. Until now, automated semantic annotation of such named entities was only applied to printed or digital text. We employ a two-step approach. First, word images are classified, identifying elements of scientific species names; Genus, species, author, using (i) visual structural features, (ii) position, and (iii) context. Second, the identified species names are semantically annotated according to the NHC-Ontology, an ontology that describes species observations. Internationalised Resource Identifiers (IRIs) are assigned to the elements so that they can be linked and disambiguated at a later stage by individual researchers. For the identification of scientific species names, we achieve an average F1 score of 0.86. Moreover, we discuss how our method will function in a semi-automated annotation process, with a fruitful dialogue between system and user as the main objective.

In Reinforcement learning, Q-learning is the best-known algorithm but it suffers from overestimation bias, which may lead to poor performance or unstable learning. In this paper, we present a novel analysis of this problem using various control tasks. For solving these tasks, Q-learning is combined with a multilayer perceptron (MLP), experience replay, and a target network. We focus our analysis on the effect of the learning rate when training the MLP. Furthermore, we examine if decaying the learning rate over time has advantages over static ones. Experiments have been performed using various maze-solving problems involving deterministic or stochastic transition functions and 2D or 3D grids and two Open-AI gym control problems. We conducted the same experiments with Double Q-learning using two MLPs with the same parameter settings, but without target networks. The results on the maze problems show that for Q-learning combined with the MLP, the overestimation occurs when higher learning rates are used and not when lower learning rates are used. The Double Q-learning variant becomes much less stable with higher learning rates and with low learning rates the overestimation bias may still occur. Overall, decaying learning rates clearly improves the performances of both Q-learning and Double Q-learning.

Q-learning is a reinforcement learning algorithm that has overestimation bias, because it learns the optimal action values by using a target that maximizes over uncertain action-value estimates. Although the overestimation bias of Q-learning is generally considered harmful, a recent study suggests that it could be either harmful or helpful depending on the reinforcement learning problem. In this paper, we propose a new Q-learning variant, called Variation-resistant Q-learning, to control and utilize estimation bias for better performance. Firstly, we present the tabular version of the algorithm and mathematically prove its convergence. Secondly, we combine the algorithm with function approximation. Finally, we present empirical results from three different experiments, in which we compared the performance of Variation-resistant Q-learning, Q-learning, and Double Q-learning. The empirical results show that Variation-resistant Q-learning can control and utilize estimation bias for better performance in the experimental tasks.

Harmony in visual compositions is a concept that cannot be defined or easily expressed mathematically, even by humans. The goal of the research described in this paper was to find a numerical representation of artistic compositions with different levels of harmony. We ask humans to rate a collection of grayscale images based on the harmony they convey. To represent the images, a set of special features were designed and extracted. By doing so, it became possible to assign objective measures to subjectively judged compositions. Given the ratings and the extracted features, we utilized machine learning algorithms to evaluate the efficiency of such representations in a harmony classification problem. The best performing model (SVM) achieved 80% accuracy in distinguishing between harmonic and disharmonic images, which reinforces the assumption that concept of harmony can be expressed in a mathematical way that can be assessed by humans., Comment: 9 pages, ICAART 2021

In the quest for efficient and robust learning methods, combining unsupervised state representation learning and reinforcement learning (RL) could offer advantages for scaling RL algorithms by providing the models with a useful inductive bias. For achieving this, an encoder is trained in an unsupervised manner with two state representation methods, a variational autoencoder and a contrastive estimator. The learned features are then fed to the actor-critic RL algorithm Proximal Policy Optimization (PPO) to learn a policy for playing Open AI's car racing environment. Hence, such procedure permits to decouple state representations from RL-controllers. For the integration of RL with unsupervised learning, we explore various designs for variational autoencoders and contrastive learning. The proposed method is compared to a deep network trained directly on pixel inputs with PPO. The results show that the proposed method performs slightly worse than directly learning from pixel inputs; however, it has a more stable learning curve, a substantial reduction of the buffer size, and requires optimizing 88% fewer parameters. These results indicate that the use of pre-trained state representations has several benefits for solving RL tasks.

This paper casts coordination of a team of robots within the framework of game theoretic learning algorithms. In particular a novel variant of fictitious play is proposed, by considering multi-model adaptive filters as a method to estimate other players’ strategies. The proposed algorithm can be used as a coordination mechanism between players when they should take decisions under uncertainty. Each player chooses an action after taking into account the actions of the other players and also the uncertainty. Uncertainty can occur either in terms of noisy observations or various types of other players. In addition, in contrast to other game-theoretic and heuristic algorithms for distributed optimisation, it is not necessary to find the optimal parameters a priori. Various parameter values can be used initially as inputs to different models. Therefore, the resulting decisions will be aggregate results of all the parameter values. Simulations are used to test the performance of the proposed methodology against other game-theoretic learning algorithms.

The practical availability of Audiovisual Processing tools to media scholars and heritage institutions remains limited, despite all the technical advancements of recent years. In this article we present the approach chosen in the CLARIAH project to increase this availability, we discuss the challenges encountered, and introduce the technical solutions we are implementing. Through three use cases focused on the enrichment of AV archives, Pose Analysis, and Automatic Speech Recognition, we demonstrate the potential and breadth of using Audiovisual Processing for archives and Digital Humanities research.

The number of bags mishandled while transferring to a connecting flight is high. Bags at-risk of missing their connections can be processed faster; however, identifying such bags at-risk is still done by simple business rules. This work researches a general model of baggage transfer process and proposes a Gradient Boosting Machine based prediction model for identifying the bags at-risk. Our prediction model is compared to the current rule based method and a benchmark using logistic regression. The results show that our model offers an increase in accuracy coupled with a marked increase in precision and recall when identifying bags that are transferred unsuccessfully.

The tiled convolutional neural network (TCNN) has been applied only to computer vision for learning invariances. We adjust its architecture to NLP to improve the extraction of the most salient features for sentiment analysis. Knowing that the major drawback of the TCNN in the NLP field is its inflexible filter structure, we propose a novel architecture called hybrid tiled convolutional neural network (HTCNN) that applies a filter only on the words that appear in similar contexts and on their neighbouring words (a necessary step for preventing the loss of some n-grams). The experiments on the IMDB movie reviews dataset demonstrate the effectiveness of the HTCNN that has a higher level of performance of more than 3% and 1% respectively than both the convolutional neural network (CNN) and the TCNN. These results are confirmed by the SemEval-2017 dataset where the recall of the HTCNN model exceeds by more than six percentage points the recall of its simple variant, CNN.

Since initial digitization strategies are often inspired by existing usage of the data, and usage of archives often varies among institutes, there is a lot of variation in accessibility of digital collections. We identify four challenges that researchers may encounter when querying such collections in conjunction for research purposes, namely query formulation, alignment, source selection and lack of transparency. We present a multiagent architecture to help overcome these challenges and discuss an prototype implementation of this framework. By means of a query scenario we show the utility of using the framework for humanities researchers.

Determining the maximum capacities of shunting yards is an important problem at Dutch Railways (NS). Solving this capacity determination problem is computational expensive as it requires to solve an NP-hard shunting planning problem. Currently, NS uses a shunt plan simulator where a local search heuristic is implemented to determine such capacities.In this paper, we study how to combine machine learning with local search in order to speed up finding shunting plans in the capacity determination problem. We investigate this in the following two ways. In the first approach, we propose to use the Deep Graph Convolutional Neural Network (DGCNN) to predict whether local search will find a feasible shunt plan given an initial solution. Using instances generated from the simulator, we build a classification model and show our approach can significantly reduce the simulation time in determining the capacity of a given shunting yard.In the second approach, we investigate whether we can use machine learning to help local search decide which promising areas to explore during search. Therefore, DGCNN is applied to predict the order of search operators in which the local search heuristic should evaluate. We show that accurately predicting the evaluation order could find improved solutions faster, and may lead to more consistent plans.

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.

High turnover of online advertising and especially real time bidding makes this ad market very attractive to beneficiary stakeholders. For publishers, it is as easy as placing some slots in their webpages and sell these slots in the available online auctions. It is important to determine which online auction market to send their slots to. Based on the traditional Waterfall Strategy, publishers have a fixed ordering of preferred online auction markets, and sell the ad slots by trying these markets sequentially. This fixed-order strategy replies heavily on the experience of publishers, and often it does not provide highest revenue. In this paper, we propose a method for dynamically deciding on the ordering of auction markets for each available ad slot. This method is based on reinforcement learning (RL) and learns the state-action through a tabular method. Since the state-action space is sparse, a prediction model is used to solve this sparsity. We analyze a real-time bidding dataset, and then show that the proposed RL method on this dataset leads to higher revenues. In addition, a sensitivity analysis is performed on the parameters of the method.

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

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

Museums, archives and digital libraries make increasing use of Semantic Web technologies to enrich and publish their collection items. The contents of those items, however, are not often enriched in the same way. Extracting named entities within historical manuscripts and disclosing the relationships between them would facilitate cultural heritage research, but it is a labour-intensive and time-consuming process, particularly for handwritten documents. It requires either automated handwriting recognition techniques, or manual annotation by domain experts before the content can be semantically structured. Different workflows have been proposed to address this problem, involving full-text transcription and named entity extraction, with results ranging from unstructured files to semantically annotated knowledge bases. Here, we detail these workflows and describe the approach we have taken to disclose historical biodiversity data, which enables the direct labelling and semantic annotation of document images in hand-written archives.

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.

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.

It is known that the comprehension of spatial prepositions involves the deployment of visual attention. For example, consider the sentence “The salt is to the left of the stove”. Researchers [29, 30] have theorized that people must shift their attention from the stove (the reference object, RO) to the salt (the located object, LO) in order to com- prehend the sentence. Such a shift was also implicitly assumed in the Attentional Vector Sum (AVS) model by [35], a cognitive model that computes an acceptability rating for a spatial preposition given a display that contains an RO and an LO. However, recent empirical findings showed that a shift from the RO to the LO is not necessary to understand a spatial preposition ( [3], see also [15, 38]). In contrast, these findings suggest that people perform a shift in the reverse direction (i.e., from the LO to the RO). Thus, we propose the reversed AVS (rAVS) model, a modified version of the AVS model in which attention shifts from the LO to the RO. We assessed the AVS and the rAVS model on the data from [35] using three model simulation methods. Our simulations show that the rAVS model performs as well as the AVS model on these data while it also integrates the recent empirical findings. Moreover, the rAVS model achieves its good performance while being less flexible than the AVS model. (This article is an updated and extended version of the paper [23] presented at the 8th International Conference on Agents and Artificial Intelligence in Rome, Italy. The authors would like to thank Holger Schultheis for helpful discussions about the additional model simulation.)

We have developed a program called mudos (maastricht university domineering solver) that solves domineering positions in a very efficient way. It enables the solution of known positions (up to the 10\times 1010\times 10 board) to be much quicker.more importantly, it enables the solution of 11\times 1111\times 11 domineering, a board size that up till now was far out of the reach of previous domineering solvers. The solution needed the investigation of 259,689,994,008 nodes, using almost half a year of computation time on a single simple desktop computer. The results show that under optimal play the first player wins 11\times 1111\times 11 domineering, irrespective whether vertical or horizontal starts.in addition, several other new boards were solved. Using the convention that vertical starts, the 8\times 158\times 15, 11\times 911\times 9, 12\times 812\times 8, 12\times 1512\times 15, 14\times 814\times 8, and 17\times 617\times 6 boards are all won by vertical, whereas the 6\times 176\times 17, 8\times 128\times 12, 9\times 119\times 11, and 11\times 1011\times 10 boards are all won by horizontal.

Regier and Carlson (2001) have investigated the processing of spatial prepositions and developed a cognitive model that formalizes how spatial prepositions are evaluated against depicted spatial relations between objects. In their Attentional Vector Sum (AVS) model, a population of vectors is weighted with visual attention, rooted at the reference object and pointing to the located object. The deviation of the vector sum from a reference direction is then used to evaluate the goodness-of-fit of the spatial preposition. Crucially, the AVS model assumes a shift of attention from the reference object to the located object. The direction of this shift has been challenged by recent psycholinguistic and neuroscientific findings. We propose a modified version of the AVS model (the rAVS model) that integrates these findings. In the rAVS model, attention shifts from the located object to the reference object in contrast to the attentional shift from the reference object to the located object implemented in the AVS model. Our model simulations show that the rAVS model accounts for both the data that inspired the AVS model and the most recent findings.

Self-Organizing Map (SOM) is a neural network model which is used to obtain a topology-preserving mapping from the (usually high dimensional) input/feature space to an output/map space of fewer dimensions (usually two or three in order to facilitate visualization). Neurons in the output space are connected with each other but this structure remains fixed throughout training and learning is achieved through the updating of neuron reference vectors in feature space. Despite the fact that growing variants of SOM overcome the fixed structure limitation they increase computational cost and also do not allow the removal of a neuron after its introduction. In this paper, a variant of SOM is proposed called AMSOM (Adaptive Moving Self-Organizing Map) that on the one hand creates a more flexible structure where neuron positions are dynamically altered during training and on the other hand tackles the drawback of having a predefined grid by allowing neuron addition and/or removal during training. Experiments using multiple literature datasets show that the proposed method improves training performance of SOM, leads to a better visualization of the input dataset and provides a framework for determining the optimal number and structure of neurons., Comment: ICAART 2016 accepted full paper

We present Contour-net, a bio-inspired model for tactile contour-tracing driven by an Hopf oscillator. By controlling the rhythmic movements of a simulated insect-like feeler, the model executes both wide searching and local sampling movements. Contour-tracing is achieved by means of contact-induced phase-forwarding of the oscillator. To classify the shape of an object, collected contact events can be directly fed into machine learning algorithms with minimal pre-processing (scaling). Three types of classifiers were evaluated, the best one being a Support Vector Machine. The likelihood of correct classification steadily increases with the number of collected contacts, enabling an incremental classification during sampling. Given a sufficiently large training data set, tactile shape recognition can be achieved in a position-, orientation- and size-invariant manner. The suitability for robotic applications is discussed.

The RoboCup simulator is a playing ground for Agents and Artificial Intelligence research. One of the main challenges provided by RoboCup is generating winning strategies for a set of agents playing soccer, given a partial and noisy view of the game state. Additionally, RoboCup is timing sensitive, and all decisions have to be sent to the server within each tick of 100ms. This paper presents a method for generating strategies by modelling players and scenarios as timed automata in the Uppaal environment. The newest extension of Uppaal, called Uppaal Stratego, allows for synthesising strategies optimising a reward function that is used to guide the decision process. In order to stay within the time frame of 100ms, two approaches were tested, namely forecasting the game state and generating a strategy asynchronously for a later point in time, and generating strategies beforehand and saving them in a lookup table. Four timed automata were developed, and were tested against publicly available methods. We found that strategies could be successfully generated and used within the time constraints of RoboCup using our proposed method. Especially when combined together, our strategies are able to outperform most published methods, but lose against the published world champions.

Voltage and current measurement data based deep learning as a method to conduct sensorless coil temperature prediction of an embedded linear induction actuator is being proposed and validated in this work. Generated numerical data from Finite Element field simulations are used to train a neural network which in turn predicts temperatures at non-accessible places e.g. at an embedded coil. The network is demonstrated and the comparison to experimental data shows the potential of virtual sensing. Even though the number of physical sensors have increased enormously in the last decades, the measurement of desired temperatures at certain locations is limited by accessibility and by the application itself, for example, if a coil is used as a moving part in an actuator. This work proposes an indirect method based on measurable quantities in the device, such as voltage and current, to quantify precisely temperatures and hot spots in sensitive parts of the device. As high temperatures can have a huge effect on the device's performance, a controllable voltage to compensate the performance reduction instantaneously is desired. Applications based on the principle of an inductive linear actuator show a strong performance dependency on the temperature of the conducting material or coil. The authors present an Artificially Intelligent voltage controller to achieve the desired performance based on measurable variables in the device and supported by sensorless methods like temperature prediction with Artificial Intelligence (AI)., Proceedings of the 13th International Conference on Agents and Artificial Intelligence, ISBN:978-989-758-484-8

This paper presents an approach for retrofitting pre-trained word representations into sense level representations to improve semantic distinction of words. We use semantic relations as positive and negative examples to refine the results of a pre-trained model instead of integrating them into the objective functions used during training. We experimentally evaluate our approach on two word similarity tasks by retrofitting six datasets generated from three widely used techniques for word representation using two different strategies. Our approach significantly and consistently outperforms three state-of-the-art retrofitting approaches.

Facial expression recognition (FER) in the context of machine learning refers to a solution whereby a computer vision system can be trained and used to automatically detect the emotion of a person from a presented facial image. FER presents a difficult image classification problem that has received increasing attention over recent years mainly due to the availability of powerful hardware for system implementation and the greater number of possible applications in everyday life. However, the FER problem has not yet been fully resolved, with the diversity of captured facial images from which the type of expression or emotion is to be detected being one of the main obstacles. Ready-made image databases have been compiled by researchers to train and test the developed FER algorithms. Most of the reported algorithms perform relatively well when trained and tested on a single-database but offer significantly inferior results when trained on one database and then tested using facial images from an entirely different database. This paper deals with the cross-database FER problem by proposing a novel approach which aggregates local region features from the eyes, nose and mouth and selects the optimal classification techniques for this specific aggregation. The conducted experiments show a substantial improvement in the recognition results when compared to similar cross-database tests reported in other works. This paper confirms the idea that, for images originating from different databases, focus should be given to specific regions while less attention is paid to the face in general and other facial sections.

The modelling of cognition is playing a major role in robotics. Indeed, robots need to learn, adapt and plan their actions in order to interact with their environment. To do so, approaches like embodiment and enactivism propose to ground sensorimotor experience in the robot's body to shape the development of cognition. In this work, we focus on the role of memory during learning in a closed loop. As sensorimotor contingencies, we consider a robot arm that moves a baby mobile toy to get visual reward. First, the robot explores the continuous sensorimotor space by associating visual stimuli to motor actions through motor babbling. After exploration, the robot uses the experience from its memory and exploits it, thus optimizing its motion to perceive more visual stimuli. The proposed approach uses Dynamic Field Theory and is integrated in the GummiArm, a 3D printed humanoid robot arm. The results indicate a higher visual neural activation after motion learning and show the benefits of an embodied babbling strategy. publishedVersion

Tuhosin alkuperäisen hyvin vajaan tietueen. Siinä oli mukana tiedosto, mutta se ei auennut. The Internet of Things (IoT) security and privacy remain a major challenge, especially due to the highly scalable and distributed nature of IoT networks. IoT security has an exceptionally wide scope which also leads to the most demanding requirements needed for the widespread realization of many IoT visions. It includes the tasks such as trusted sensing, computation, communication, privacy in terms of security scope. A large number of industry-driven domain specific standards has been developed which hinder the development of a single IoT ecosystem. O-MI and O-DF standards were previously proposed to address the challenge of interoperability in IoT devices. This paper highlights and discusses the measures taken to enhance the security of these standards by integration of OAuth 2.0 service provider functionality with the O-MI authentication module along with the existing security module. The security architecture has been proposed which enhances the security features by providing authorization without use of certificates. The implementation details of the plug-in module for envisioned security model developed for the O-MI and O-DF standards have been discussed.

Many years after online social networks exceeded our collective attention, social influence is still built on attention capital. Quality is not a prerequisite for viral spreading, yet large diffusion cascades remain the hallmark of a social influencer. Consequently, our exposure to low-quality content and questionable influence is expected to increase. Since the conception of influence maximization frameworks, multiple content performance metrics became available, albeit raising the complexity of influence analysis. In this paper, we examine and consolidate a diverse set of content engagement metrics. The correlations discovered lead us to propose a new, more holistic, one-dimensional engagement signal. We then show it is more predictable than any individual influence predictors previously investigated. Our proposed model achieves strong engagement ranking performance and is the first to explain half of the variance with features available early. We share the detailed numerical workflow to compute the new compound engagement signal. The model is immediately applicable to social media monitoring, influencer identification, campaign engagement forecasting, and curating user feeds., Presented at ICAART 2020

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

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

Cardio Vascular Disease (CVD) also known as heart and circulatory disease comprises all the illnesses of the heart and the circulatory system, namely coronary heart disease, angina, heart attack, congenital heart disease or stroke. CVDs are, nowadays, one of the main causes of death. Indeed, this fact reveals the centrality of prevention and how important is to be aware on these kind of situations. Thus, this work will focus on the development of a decision support system to help to prevent these events from happening, centred on a formal framework based on Mathematical Logic and Logic Programming for Knowledge Representation and Reasoning, complemented with a Case Based Reasoning approach to computing that caters to the handling of incomplete, unknown or even self-contradictory information or knowledge., This work has been supported by COMPETE: POCI01-0145-FEDER-007043 and FCT – Fundação para a Ciência e Tecnologia within the Project Scope: UID/CEC/00319/2013.

The ability of an autonomous agent to select rational actions is vital in enabling it to achieve its goals. To do so effectively in a high-stakes setting, the agent must be capable of considering the risk and potential reward of both immediate and future actions. In this paper we provide a novel method for calculating risk alongside utility in online planning algorithms. We integrate such a risk-aware planner with a BDI agent, allowing us to build agents that can set their risk aversion levels dynamically based on their changing beliefs about the environment. To guide the design of a risk-aware agent we propose a number of principles which such an agent should adhere to and show how our proposed framework satisfies these principles. Finally, we evaluate our approach and demonstrate that a dynamically risk-averse agent is capable of achieving a higher success rate than an agent that ignores risk, while obtaining a higher utility than an agent with a static risk attitude.

These transactions publish research in computer-based methods of computational collective intelligence (CCI) and their applications in a wide range of fields such as the semantic Web, social networks, and multi-agent systems. TCCI strives to cover new methodological, theoretical and practical aspects of CCI understood as the form of intelligence that emerges from the collaboration and competition of many individuals (artificial and/or natural). The application of multiple computational intelligence technologies, such as fuzzy systems, evolutionary computation, neural systems, consensus theory, etc., aims to support human and other collective intelligence and to create new forms of CCI in natural and/or artificial systems. This twentieth issue contains 11 carefully selected and revised contributions.

The knowledge acquired through past experiences is of the most importance when humans or machines try to find solutions for new problems based on past ones, which makes the core of any Case-based Reasoning approach to problem solving. On the other hand, existent CBR systems are neither complete nor adaptable to specific domains. Indeed, the effort to adapt either the reasoning process or the knowledge representation mechanism to a new problem is too high, i.e., it is extremely difficult to adapt the input to the computational framework in order to get a solution to a particular problem. This is the drawback that is addressed in this work., This work is funded by National Funds through the FCT – Fundação para a Ciência e a Tecnologia (Portuguese Foundation for Science and Technology) within projects PEst-OE/EEI/UI0752/2014 and PEst-OE/QUI/UI0619/2012.