Catálogo de publicaciones - libros

Consider a robot whose task is to pick up some colored balls from a grid, taking the red balls to a red spot, the blue balls to a blue spot and so on, one by one, without knowing either the location or color of the balls but having a sensor that can find out both when a ball is near. This problem is simple and can be solved by a domain-independent contingent planner in principle, but in practice this is not possible: the size of any valid plan constructed by a contingent planner is exponential in the number of observations which in these problems is very large. This doesn’t mean that planning techniques are of no use for these problems but that building or verifying complete contingent plans is not feasible in general. In this work, we develop a domain-independent action selection mechanism that does not build full contingent plans but just chooses the action to do next in a closed-loop fashion. For this to work, however, the mechanism must be both fast and informed. We take advantage of recent ideas that allow delete and precondition-free contingent problems to be converted into conformant problems, and conformant problems into classical ones, for mapping the action selection problem in contingent planning into an action selection problem in classical planning that takes sensing actions into account. The formulation is tested over standard contingent planning benchmarks and problems that require plans of exponential size.

This paper explores an integrated approach to diagnosis of complex dynamic systems. Consistency-based diagnosis is capable of performing automatic fault detection and localization using just correct behaviour models. Nevertheless, it may exhibit low discriminative power among fault candidates. Hence, we combined the consistency based approach with machine learning techniques specially developed for fault identification of dynamic systems. In this work, we apply Stacking to generate time series classifiers from classifiers of its univariate time series components. The Stacking scheme proposed uses K-NN with Dynamic Time Warping as a dissimilarity measure for the level 0 learners and Naïve Bayes at level 1. The method has been tested in a fault identification problem for a laboratory scale continuous process plant. Experimental results show that, for the available data set, the former Stacking configuration is quite competitive, compare to other methods like tree induction, Support Vector Machines or even K-NN and Naïve Bayes as stand alone methods.

This paper addresses the characterization of a large text collection by introducing a method for retrieving sets of relevant WordNet concepts as descriptors of the collection contents. The method combines models for identifying interesting word co-occurrences with an extension of a word sense disambiguation algorithm in order to retrieve the concepts that better fit in with the collection topics. Multi-word nominal concepts that do not explicitly appear in the texts, can be found among the retrieved concepts. We evaluate our proposal using extensions of recall and precision that are also introduced in this paper.

Collaborative work can be supported by many tools and it has been included in a large number of learning environments design. This paper presents issues related to an educational portal design and collaboration in Intelligent Tutoring Systems (ITS). In order to achieve the collaboration it was necessary to provide a way to interoperate knowledge among the heterogeneous systems. We have been developing ITS as resources to improve the individual and personalized learning. We believe that individual experiences can be more successful when the student has more autonomy and he is less dependent of the professor. In this research direction, this paper details the Social Agent reasoning, an agent to improve student’s learning stimulating his interaction with other students, and how this agent exchange bayesian knowledge among AMPLIA agents. The AMPLIA environment is an Intelligent Probabilistic Multi-agent Environment to support the diagnostic reasoning development and the diagnostic hypotheses modeling of domains with complex and uncertain knowledge, like medical area.

This paper describes an approach to automatically obtain an HTN planning domain from a well structured learning objects repository and also to apply an HTN planner to obtain IMS Learning Designs adapted to the features and needs of every student.

Generative Topographic Mapping is a probabilistic model for data clustering and visualization. It maps points, considered as prototype representatives of data clusters, from a low dimensional latent space onto the observed data space. In semi-supervised settings, class information can be added resulting in a model variation called class-GTM. The number of class-GTM latent points used is usually large for visualization purposes and does not necessarily reflect the class structure of the data. It is therefore convenient to group the clusters further in a two-stage procedure. In this paper, class-GTM is first used to obtain the basic cluster prototypes. Two novel methods are proposed to use this information as prior knowledge for the K-means-based second stage. We evaluate, using an entropy measure, whether these methods retain the class separability capabilities of class-GTM in the two-stage process, and whether the two-stage procedure improves on the direct clustering of the data using K-means.

This article presents an approach for noise filtering that is based on anisotropic nonlinear diffusion. The method combines edge-preserving noise reduction with a strategy to enhance local structures and a mechanism to further smooth the background. We have provided the method with an automatic mechanism for parameter self-tuning and for stopping the iterative filtering process. The performance of the approach is illustrated with its application to electron cryotomography (cryoET). CryoET has emerged as a leading imaging technique for visualizing the molecular architecture of complex biological specimens. A challenging computational task in this discipline is to increase the extremely low signal-to-noise ratio (SNR) to allow visualization and interpretation of the three-dimensional structures. The filtering method here proposed succeeds in substantially reducing the noise with excellent preservation of the structures.

This work presents a middle-ware able to translate web services descriptions into a temporal HTN domain in order to automatically compose and execute sequences of web service invocations, including parallel branches and complex synchronizations, based on the combination of HTN planning and temporal reasoning techniques.

We consider a job shop problem with uncertain durations and flexible due dates and introduce a multiobjective model based on lexicographical minimisation. To solve the resulting problem, a genetic algorithm and a decoding algorithm to generate possibly active schedules are considered. The multiobjective approach is tested on several problem instances, illustrating the potential of the proposed method.

Being aware of the importance of classifiers to be comprehensible when using machine learning to solve real world problems, bagging needs a way to be explained. This work compares Consolidated Tree’s Construction (CTC) algorithm with the Combined Multiple Models (CMM) method proposed by Domingos when used to extract explanation of the classification made by bagging. The comparison has been done from two main points of view: accuracy, and quality of the provided explanation. From the experimental results we can conclude that it is recommendable the use of CTC rather than the use of CMM. From the accuracy point of view, the behaviour of CTC is nearer the behaviour of bagging than CMM’s one. And, analysing the complexity of the obtained classifiers, we can say that Consolidated Trees (CT trees) will give simpler and, therefore, more comprehensible explanation than CMM classifiers. And besides, looking to the stability of the structure of the built trees, we could say that the explanation given by CT trees is steadier than the one given by CMM classifiers. As a consequence, the user of the classifier will feel more confident using CTC than using CMM.