loading subjects...
Ebook: Artificial Intelligence Research and Development
loading...
This book presents 34 original papers accepted for presentation at the 17th International Conference of the Catalan Association for Artificial Intelligence (CCIA 2014), held in October 2014 in Barcelona, Spain.
The Catalan Association for Artificial Intelligence (ACIA), was created in 1994 as a non-profit association to promote cooperation among researchers from the Catalan-speaking artificial intelligence research community. Conferences are now held annually throughout the Catalan-speaking countries.
The papers in this volume have been organized around different topics, providing a representative sample of the current state-of-the-art in the Catalan artificial intelligence community and of the collaboration between ACIA members and the worldwide AI community. The book will be of interest to all those working in the field of artificial intelligence.
The Catalan Association for Artificial Intelligence (ACIA),
ACIA is a member of ECCAI, the European Coordinating Committe for Artificial Intelligence (http://www.acia.org).
This year the conference is being held in Barcelona, hosted by the Universitat de Barcelona, and organised by the ACIA together with the Applied Mathematics and Analysis department of the hosting university. CCIA 2014 has been organised as a single-track conference consisting of high quality, previously unpublished papers on new and original research on Artificial Intelligence.
This volume contains 34 original contributions, which have been accepted for presentation at the Seventeenth International Conference of the Catalan Association of Artificial Intelligence (CCIA 2014), which will take place on October 22–24, 2014. All contributions have been reviewed by at least two referees. The papers have been organised around different topics providing a representative sample of the current state of the art in the Catalan Artificial Intelligence Community and of the collaboration between ACIA members and the worldwide AI community.
As a novelty, this year conference hosts the Cognitive Science Society (CSS)
http://cognitivesciencesociety.org
We would like to express our sincere gratitude to all authors for making the conference and this book possible with their contributions and participation. We sincerely thank the members of the organizing committee for their effort in the preparation of this event, and all the members of the Scientific Committee for their help in the reviewing process.
Special thanks go also to the two outstanding plenary speakers, Ramón López de Mántaras and Leo Wanner, for their effort in preparing very interesting lectures, to Karina Gibert, former PC chair in CCIA'2013, for her help during this year and, last but not least, to Vicenç Torra, president of ACIA, for his kind support and involvement.
We wish all participants a successful and inspiring conference and a pleasant stay in Barcelona.
Lledó Museros, Universitat Jaume I
Oriol Pujol, Universitat de Barcelona
Núria Agell, ESADE Business School
October 2014
Reinforcement Learning (RL) provides a general methodology to solve complex uncertain decision problems, which are very challenging in many real-world applications. RL problem is modeled as a Markov Decision Process (MDP) deeply studied in the literature. We consider Policy Iteration (PI) algorithms for RL which iteratively evaluate and improve control policies. In handling problems with continuous states or in very large state spaces, generalization is mandatory. Generalization property of RL algorithms is an important factor to predict values for unexplored states. Candidates for value function approximation are Support Vector Regression (SVR) known to have good properties over the generalization ability. SVR has been used in batch frameworks in RL but, smart implementations of incremental exact SVR can extend SVR generalization ability to online RL where the expected reward from states change constantly with experience. Hence our online SVR is a novelty method which allows fast and good estimation of value function achieving RL objective very efficiently. Throughout simulation tests, the feasibility and usefulness of the proposed approach is demonstrated.
Random sampling is an efficient method dealing with constrained optimization problems. In computational geometry, it has been applied, through Clarkson's algorithm [10], to solve a general class of problems called violator spaces. In machine learning, TSVM is a learning method used when only a small fraction of labeled data is available, which implies solving a non convex optimization problem. Several approximation methods have been proposed to solve it, but they usually find suboptimal solutions. Global optimal solution may be obtained using exact techniques, costing an exponential time complexity with respect to the number of instances. In this paper, an interpretation of TSVM in terms of violator space is given. Hence, a randomized method is presented extending the use of exact methods now reducing the time complexity to sub-exponential in particular exponential w.r.t. the number of support vectors of the optimal solution instead of exponential w.r.t. the number of instances.
The last decades have shown an increasing interest in studying how to automatically capture the likeness or proximity among data objects due to its importance in machine learning and pattern recognition. Under this scope, two major approaches have been followed that use either feature-based or distance-based representations to perform learning and classification tasks. This paper presents the first results of a comparative experimental study between these two approaches for computing similarity scores using a classification-based method. In particular, we use the Support Vector Machine, as a flexible combiner both for a high dimensional feature space and for a family of distance measures, to finally learn similarity scores in a CBIR context. We analyze both the influence of the different input data formats and the training size on the performance of the classifier. Then, we found that a low dimensional multidistance-based representation can be convenient for small to medium-size training sets whereas it is detrimental as the training size grows.
By increasing popularity of wearable cameras, life-logging data analysis is becoming more and more important and useful to derive significant events out of this substantial collection of images. In this study, we introduce a new tracking method applied to visual life-logging, called bag-of-tracklets, which is based on detecting, localizing and tracking of people. Given the low spatial and temporal resolution of the image data, our model generates and groups tracklets in a unsupervised framework and extracts image sequences of person appearance according to a similarity score of the bag-of-tracklets. The model output is a meaningful sequence of events expressing human appearance and tracking them in life-logging data. The achieved results prove the robustness of our model in terms of efficiency and accuracy despite the low spatial and temporal resolution of the data.
This paper presents a research in progress towards autonomous underwater robot manipulation. Current research in underwater robotics intends to increase the autonomy of intervention operations that require physical interaction. Autonomous grasping is still a very challenging skill, especially in underwater environments, with highly unstructured scenarios, limited availability of sensors and adverse conditions that affect the robot perception and control systems in various degrees. To tackle those issues, we propose the use of vision and segmentation techniques that aim to improve the specification of grasping operations on underwater primitive shaped objects. Several sources of stereo information are used to gather 3D information in order to obtain a model of the object. Using a RANSAC primitive shape recognition algorithm, the model parameters are estimated and a set of feasible grasps are computed. This approach is validated in simulation and the quality of different 3D reconstructions from both real and virtual scenarios is analyzed.
In this work we build a system for automatic emotion classification from image sequences. We analyze subtle changes in facial expressions by detecting a subset of 12 representative facial action units (AUs). Then, we classify emotions based on the output of these AUs classifiers, i.e. the presence/absence of AUs. We base the AUs classification upon a set of spatio-temporal geometric and appearance features for facial representation, fusing them within the emotion classifier. A decision tree is trained for emotion classifying, making the resulting model easy to interpret by capturing the combination of AUs activation that lead to a particular emotion. For Cohn-Kanade database, the proposed system classifies 7 emotions with a mean accuracy of near 90%, attaining a similar recognition accuracy in comparison with non-interpretable models that are not based in AUs detection.
The use of conventional water resources for golf course activities is increasingly contested. As a result, in recent years there has been a considerable demand for golf courses to adopt environmentally sustainable strategies. Moreover, other exploitation-derived resources of turfgrass treatment have important economic costs. Thus, a major optimisation in resources and time is required, which is covered by the WATERGOLF project through different modules. This paper presents an intelligent system to optimise water consumption in golf courses, as well as to improve their management. Hence, the system is designed to suggest corrective actions through an expert system to the greenkeeper, as well as to trigger alarms in risky situations in four different modules: irrigation, weeds, diseases, and fertility. These suggestions are based on expert knowledge acquired from turfgrass experts and a set of measurements gathered by advanced sensors placed in the golf course. Although the intelligent system is currently being developed, it is expected that a further evaluation in real conditions will help golf facilities to reduce their maintenance costs and environmental impact.
This paper presents a study of the applicability and suitability of ELECTRE-III-H for evaluating sectorial water allocation policies for a water stressed Mediterranean area of Northeastern Spain (Tarragona). Proposed method is based on the outranking model of decision support systems which constructs a partial preorder based on pairwise preference relations among all the possible actions. This work has addressed the multi-criteria water management problem considering several water allocation strategies to mitigate water scarcity induced by climate change. We compare several adaptation measures including alternative water sources, inter basin water transfer and sectorial demand management coming from industry, agriculture and domestic sectors.
In this paper, a new formulation of the central ideas of the well-established theory of Boden about creativity is presented. This new formulation redefines some terms and reviews the formal mechanisms of exploratory and transformational creativity. The presented approach is based on the conceptual space proposed by Boden and formalized by other authors in a way that facilitates the implementation of these mechanisms. The presented formulation is applied to a the real case of creative designing in which a new combination of chocolate and fruit is desired. The experimentation has been conducted jointly with a Spanish chocolate chef. Data collected from the chef has been used to validate the proposed system. Experimental results show that the formulation presented is not only useful for understanding how the creative mechanisms of design works, but also facilitates its implementation in real cases to support creativity processes.
Performing subsea intervention tasks is a challenge due to the complexities of the underwater domain. We propose to use a learning by demonstraition algorithm to intuitively teach an intervention autonomous underwater vehicle (I-AUV) how to perform a given task. Taking as an input few operator demonstrations, the algorithm generalizes the task into a model and simultaneously controls the vehicle and the manipulator (using 8 degrees of freedom) to reproduce the task. A complete framework has been implemented in order to integrate the LbD algorithm with the different onboard sensors and actuators. A valve turning intervention task is used to validate the full framework through real experiments conducted in a water tank.
In this paper we explore the possibility of capturing color trends and understanding the rationale behind the popularity of a color. To this end, we propose using a preference disaggregation approach from the field of Multi-Criteria Decision Analysis. The main objective is to identify the criteria aggregation model that underlies the global preference of a color. We introduce a new disaggregation method based on the well-known UTASTAR algorithm able to represent preferences by means of non-monotonic utility functions. The method is applied to a large database of ranked colors, from three different years, based on the information published on the webpage of an international creative community. Non-monotone marginal utility functions from each of the coordinates are obtained for each year. These functions contain the color preference information captured, in an understandable way.
To date, recommendation systems have focused mainly on recommending products to individual rather than groups of people intending to participate in a group activity. In the last decade, with the growth of interactive activities over the internet such as e-commerce services or social virtual spaces, there has appeared many recommendation scenarios that involve groups of inter-related users. Though there have been attempts to establish group recommendation, most of them focus on off-line environments. In this paper we present a novel web-based environment that supports on-line group recommendation scenarios. Specifically, we propose a Collaborative Advisory CHannel for group recommendation called gCOACH. We introduce the environment with the multiple interaction modalities developed to communicate, coordinate and persuade the group in a case-based group recommender. We demonstrate its usability through a live-user case-study.
Location Based Social Networks (LBSN) have become an interesting source for mining user behavior. These networks (e.g. Twitter, Instagram or Foursquare) collect spatio-temporal data from users in a way that they can be seen as a set of collective and distributed sensors on a geographical area. Processing this information in different ways could result in patterns useful for several application domains. These patterns include simple or complex user visits to places in a city or groups of users that can be described by a common behavior. The domains of application range from the recommendation of points of interest to visit and route planning for touristic recommender systems to city analysis and planning.
This paper presents the analysis of data collected for several months from such LBSN inside the geographical area of two large cities. The goal is to obtain by means of unsupervised data mining methods sets of patterns that describe groups of users in terms of routes, mobility patterns and behavior profiles that can be useful for city analysis and mobility decisions.
In this paper we introduce an automated assessment service for online learning support in the context of communities of learners. The goal is to introduce automatic tools to support the task of assessing massive number of students as needed in Massive Open Online Courses (MOOC). The final assessments are a combination of tutor's assessment and peer assessment. We build a trust graph over the referees and use it to compute weights for the assesments aggregations.
Breast cancer is one of the most common neoplasms in women and it is a leading cause of worldwide death. However, it is also among the most curable cancer types if it can be diagnosed early through a proper mammographic screening procedure. So, suitable computer aided detection systems can help the radiologists to detect many subtle signs, normally missed during the first visual examination. This study proposes a Gabor filtering method for the extraction of textural features by multi-sized evaluation windows applied to the four probabilistic distribution moments. Then, an adaptive strategy for data selection is used to eliminate the most irrelevant pixels. Finally, a pixel-based classification step is applied by using Support Vector Machines in order to identify the tumor pixels. During this part we also estimate the appropriate kernel parameters to obtain an accurate configuration for the four existing kernels. Experiments have been conducted on different training-test partitions of mini-MIAS database, which is commonly used among researchers who apply machine learning methods for breast cancer diagnosis. The improved performance of our framework is evaluated using several measures: classification accuracy, positive and negative predictive values, receiver operating characteristic curves and confusion matrix.
In this paper we analyse the performance of various texture analysis methods for the purpose of breast mass detection. We considered well-known methods such as local binary patterns, histogram of oriented gradients, cooccurrence matrix features and Gabor filters. Moreover, we propose the use of local directional number patterns as a new feature extraction method for breast mass detection. For each method, a Support Vector Machine is trained on the extracted features to predict the class (mass/normal) of unknown instances. In order to improve the mass detection capability of each individual method we used classifier majority voting and feature combination techniques. Some experiments were performed on the images obtained from a public breast cancer database, achieving promising levels of sensitivity and specificity.
The problem of colour naming consists of successfully recognizing, categorising and labeling colours. This paper presents a colour naming theory based on a Qualitative Colour Description (QCD) model, which is validated here by an experiment carried out by real users in order to determine whether the QCD model is close enough to common human colour understanding. Then, this paper presents the necessity to customize this QCD by means of a user profile, since different users may perceive colours differently because of different cultures, knowledge, experiences and even physiology. A standard QCD model adaptable to users could improve colour reference and grounding in human-machine communication situations.
This paper studies T-preorders by using their Representation Theorem that states that every T-preorder on a set X can be generated in a natural way by a family of fuzzy subsets of X. Especial emphasis is made on the study of one-dimensional T-preorders (i.e.: T-preorders that can be generated by only one fuzzy subset). Strong complete T-preorders are characterized.