Catálogo de publicaciones - libros

PolyTope ARTMAP (PTAM) [6] is an ART neural network based on internal categories with irregular polytope (polygon in IR) geometry. Categories in PTAM do not overlap, so that their expansion is limited by the other categories, and not by the category size. This makes the vigilance parameter unnecessary. What happens if categories have irregular geometries but overlap is allowed? This paper presents Overlapping PTAM (OPTAM), an alternative to PTAM based on polytope overlapping categories, which tries to answer this question. The comparison between the two approaches in classification tasks shows that category overlap does not reduce neither the classification error nor the number of categories, and it also requires vigilance as a tuning parameter. Futhermore, OPTAM provides a significant variability in the results among different data sets.

The Electronic Government is a new field of applications for the semantic web where ontologies are becoming an important research technology. The e-Government faces considerable challenges to achieve interoperability given the semantic differences of interpretation, complexity and width of scope. In this paper we present the results obtained in an ongoing project commissioned by the Spanish government that seeks strategies for the e-Government to reduce the problems encountered when delivering services to citizens. We also introduce an e-Government ontology model; within this model a set of legal ontologies are devoted to representing the Real-estate transaction domain used to illustrate this paper.

Conformant planning is a variation of classical AI planning where the initial state is partially known and actions can have non-deterministic effects. While a classical plan must achieve the goal from a given initial state using deterministic actions, a conformant plan must achieve the goal in the presence of uncertainty in the initial state and action effects. Conformant planning is computationally harder than classical planning, and unlike classical planning, cannot be reduced polynomially to SAT (unless P = NP). Current SAT approaches to conformant planning, such as those considered by Giunchiglia and colleagues, thus follow a generate-and-test strategy: the models of the theory are generated one by one using a SAT solver (assuming a given planning horizon), and from each such model, a candidate conformant plan is extracted and tested for validity using another SAT call. This works well when the theory has few candidate plans and models, but otherwise is too inefficient. In this paper we propose a different use of a SAT engine where conformant plans are computed by means of This transformed theory is obtained by the original theory over the action variables. This operation, while intractable, can be done efficiently provided that the original theory is compiled into d– (Darwiche 2001), a form akin to s (Bryant 1992). The experiments that are reported show that the resulting planner is competitive with state-of-the-art optimal conformant planners and improves upon a planner recently reported at ICAPS-05.

This paper presents an architecture that makes it possible to construct dynamic systems capable of growing in dimension and adapting its knowledge to environmental changes. An architecture must define the components of the system (agents in this case), as well as the way in which those components communicate and interact with each other in order to achieve the system’s goals. The work presented here focuses on the development of an agent-based architecture, based on the use of deliberative agents, that incorporate case based reasoning. The proposed architecture requires an analysis and design methodology that facilitates the building of distributed systems using this technology. The proposal combines elements of existing methodologies such as Gaia and AUML in order to take advantage of their characteristics. Moreover the architecture takes into account the possibility of modelling problems in dynamic environments and therefore the use of autonomous models that evolve over time. To solve this problem the architecture incorporates CBR-agents whose aim is to acquire knowledge and adapt themselves to environmental changes. The architecture has been applied to model for evaluating the interaction between the atmosphere and the ocean, as well as for the planification and optimization of sea routes for vessels. The system has been tested successfully, and the results obtained are presented in this paper.

This document presents an analysis of Music Knowledge as a first step towards music representation for composition. After an introductory review of music computing evolution, several approaches to music knowledge are described: the system levels context, music theory and disciplines, dimensions in music, and finally the creative process. Then, the composition knowledge is analyzed at the symbolic level, dissecting its sub-level structure, and concluding with some requirements for an efficient representation. EV meta-model is presented as a multilevel representation tool for event based systems as music. Its structure and unique features are described within the analyzed level context. Three musical application examples of EV modeling are shown in the field of sound synthesis and music composition. These examples test representation, extension and development features.

An image can be decomposed into different elementary descriptors depending on the observer interest. Similar techniques as used to understand words, regarded as molecules, formed by combining atoms, are proposed to describe images based on their information content. In this paper, we use primitive feature extraction and clustering to code the image information content. Our purpose is to describe the complexity of the information based on the combinational profile of the clustered primitive features using entropic measures like mutual information and Kullback-Leibler divergence. The developed method is demonstrated to asses image complexity for further applications to improve Earth Observation image analysis for sustainable humanitarian crisis response in risk reduction.

In this paper, a general technique for automatically defining multilayer Cellular Neural Networks to perform Chebyshev optimal piecewise linear approximations of nonlinear functions is proposed. First, a novel CNN cell output function is proposed. Its main goal is to control input and output dynamic ranges. Afterwards, this 2-layer CNN is further programmed to achieve generic piecewise Chevyshev polynomials that approximate a nonlinear function.

Accurate shot boundary detection techniques have been an important research topic in the last decade. Such interest is motivated by the fact that segmentation of a video stream is the first step towards video content analysis and content-based video browsing and retrieval. In this paper, we present a new algorithm mainly focused on the detection of fades using a non-common feature in previous work: entropy, a scalar representation of the amount of information conveyed by each video frame. A survey of the properties of this feature is first provided, where authors show that the pattern this series exhibits when fades occur is clear and consistent. It does not depend on the monochrome color used to fade and, in addition, it is able to deal with on-screen text that sometimes remain in the monochrome stage. A statistical model-based algorithm to detect fades is proposed. Due to the clear pattern shown by fades in the entropy series and the accurate mathematical model used, motion and illumination changes do not severely affect precision as it normally happens with algorithms dealing with the detection of gradual transitions.

In this paper, we enrich the logic of order of magnitude qualitative reasoning by means of a new notion of negligibility which has very useful properties with respect to operations of real numbers. A complete axiom system is presented for the proposed logic, and the new negligibility relation is compared with previous ones and its advantages are presented on the basis of an example.

We enhance real-time search algorithms with bounded propagation of heuristic changes. When the heuristic of the current state is updated, this change is propagated consistently up to states. Applying this idea to HLRTA*, we have developed the new HLRTA*() algorithm, which shows a clear performance improvement over HLRTA*. Experimentally, HLRTA*() converges in less trials than LRTA*(), while the contrary was true for these algorithms without propagation. We provide empirical results showing the benefits of our approach.