Catálogo de publicaciones - libros

This thesis proposal aims to provide a new approach to the study of complex adaptive systems in social sciences through a methodological framework for modeling and simulating these systems like artificial societies. Agent based modeling (ABM) is well fitted for the study of social systems as it focuses on how local interactions among agents generate emergent larger and global social structures and patterns of behavior. The issues addressed by our framework are presented as well as its most important components.

When building natural language generation applications it is desireable to have the possibility of assembling modules that use different techniques for each one of the specific generation tasks. This paper presents an agent-based module for referring expression generation and aggregation, implemented within the framework of a generic architecture for implementing multi-agent systems: Open Agent Architecture.

This paper presents a system for hand posture recognition that works with colour video streams under varying light conditions for human-machine interaction in vision-based interface tasks. No initialization of the system is required and no user dependence is involved. With this aim, we first model on-line each user’s skin colour from the skin cue imaging of his/her face detected by means of Viola and Jones detector. Afterwards, a second order isomorphism approach performs tracking on skin colour blob based detected hand. Also, we propose this approach as a mechanism to estimate hand transition states. Finally, evidences about hand postures are recognized by shape matching, which is carried out through a holistic similarity measure focused on the Hausdorff distance. The paper includes experimental evaluations of the recognition system for 16 different hand postures in different video streams. The results show that the system can be suitable for real-time interfaces using general purpose hardware.

In this paper we present a concise analysis of the requirements for effective action selection, and a centralized action selection model that fulfills most of these requirements. In this model, action selection occurs by combining sensory information from the non-homogenous sensors of an off-the-shelf robot with the feedback from competing behavioral modules. In order to successfully clean an arena, the animal robot () has to present a coherent overall behavior pattern for both appropriate selection and termination of a selected behavior type. In the same way, an animat set in a chasing task has to present opportunist action selection to locate the nearest target. In consequence, both an appropriate switching of behavior patterns and a coherent overall behavior pattern are necessary for effective action selection.

This paper presents a method using intervals for representing and synthesizing imprecise information for multi-attribute evaluation and decision-making support. An implementation is given for selecting an alternative for a project in a real case in the context of construction in civil engineering. As a previous step to aggregate the available information, a methodology is proposed for summarizing and normalizing values in an intervals context, representing the alternatives by means of rectangles in (products of finite closed intervals in ). A distance is introduced in the set of rectangles, defining a total order once a reference rectangle is considered. A method is given for the choice of the best alternative based on the comparison of distances to a reference rectangle. The constraints which guarantee consistency are determined and the consistency of the method is established.

This paper analyzes how to introduce machine learning algorithms into the process of direct volume rendering. A conceptual framework for the optical property function elicitation process is proposed and particularized for the use of attribute-value classifiers. The process is evaluated in terms of accuracy and speed using four different off-the-shelf classifiers (J48, Naïve Bayes, Simple Logistic and ECOC-Adaboost). The empirical results confirm the classification of biomedical datasets as a tough problem where an opportunity for further research emerges.

P2P file sharing systems are distributed systems consisting of interconnected nodes able to organize themselves in networks, with the purpose of sharing content. Recent empirical studies have shown that they suffer from freeloaders, that is, peers that consume many more resources or content than they contribute. In this paper we propose a coalition formation based incentive mechanism for P2P file sharing systems, that improves the system performance for the coalition participant peers. In addition, it discourages free-loader like behavior. The mechanism presents a formal approach to the problem based on game theory that takes into account the rational and self-interested behavior of the peers.

Automatic detection systems do not perform as well as human observers, even on simple detection tasks. A potential solution to this problem is training vision systems on appropriate regions of interests (ROIs), in contrast to training on predefined and arbitrarily selected regions. Here we focus on detecting pedestrians in static scenes. Our aim is to answer the following question:

A method for obtaining an approximate solution to the inverse kinematics of a articulated chain is proposed in this paper. Specifically, the method is applied to determine the joint positions of a humanoid robot in locomotion tasks, defining the successive stable robot configurations needed to achieve the final foot position in each step. Our approach is based on the postural scheme method, using artificial neural networks to solve the problem. In this paper we define the restrictions that must be accomplished by the networks and make an exhaustive study about the learning algorithms, transfer functions, training sets composition, data normalization and artificial neural network topologies.

The paper reconsiders the importance of monotonicity and consistency properties on the efficiency of multiobjective A search. Previous works on the MOA algorithm (Multi-objective A) concluded that the importance of the monotone property of heuristics was not as important as in A. The recent development of an alternative algorithm (NAMOA), gives a chance to review these results. The paper presents a formal analysis on the comparison of heuristics in NAMOA and concludes that the properties of consistency and monotonicity are of fundamental importance in search efficiency.