Catálogo de publicaciones - libros

In this paper we compare, by means of numerical simulations, seven different methods for reconstructing incomplete fuzzy preference relations. We consider the case of highly inconsistent preference relations as well as the case of preference relations close to consistency. We compare the numerical results on the basis of the consistency of the reconstructed preference relations.

Web personalization is the process of customizing a Web site to the preferences of users, according to the knowledge gained from usage data in the form of user profiles. In this work, we experimentally evaluate a fuzzy clustering approach for the discovery of usage profiles that can be effective in Web personalization. The approach derives profiles in the form of clusters extracted from preprocessed Web usage data. The use of a fuzzy clustering algorithm enable the generation of overlapping clusters that can capture the uncertainty among Web users navigation behavior based on their interest. Preliminary experimental results are presented to show the clusters generated by mining the access log data of a Web site.

In this paper we explore the application of fuzzy logic and data fusion techniques to improve the performance of passage retrieval in open domain Question Answering Systems (QAS) . Our experiments show that our proposed mechanisms provide significant performance improvements when compared to other similar systems.

This paper presents a first step toward the formalization of the concept of document reliability in the context of Information Retrieval (and Information Filtering). Our proposal is based on the hypothesis that the evaluation of the relevance of a document can also depend on the concept of reliability of a document. This concept has the following properties: (i) it is user-dependent, i.e., a document may be reliable for a user and not reliable for another user; (ii) it is source-dependent, i.e., the source which a document comes from may influence its reliability for a user; and (iii) it is also author-dependent, i.e., the information about who wrote the document may also influence the user when assessing the reliability of a document.

The conceptual formalism supported by an ontology is not sufficient for handling vague information that is commonly found in many application domains. We describe how to introduce fuzziness in an ontology. To this aim we define a framework consisting of a fuzzy ontology based on Fuzzy Description Logic and Fuzzy–Owl.

In this paper, we suggest an improved weight decision rule depending on SNNR (Signal Plus Noise to Noise Ratio) and fuzzy value for simultaneous multi-modality including a synchronization between audio-gesture modalities. In order to insure the validity of the suggested weight decision rule, we implement a wireless PDA-based Multi-Modal Fusion Architecture (hereinafter, MMFA) by coupling embedded speech and KSSL recognizer, which fuses and recognizes 130 word-based instruction models that are represented by speech and KSSL (Korean Standard Sign Language), and then translates recognition result into synthetic speech (TTS) and visual illustration in real-time. In the experimental results, the average recognition rate of the MMFA fusing 2 sensory channels based on wireless PDA was 96.54% in clean environments (e.g. office space), and 93.21% in noisy environments, with the 130 word-based instruction models.

How to protect the copyright of digital media over the Internet is a problem for the creator/owner. A novel support vector regression (SVR) based digital audio watermarking scheme in the wavelet domain which using subsampling is proposed in this paper. The audio signal is subsampled and all the sub-audios are decomposed into the wavelet domain respectively. Then the watermark information is embedded into the low-frequency region of random one sub-audio. With the high correlation among the sub-audios, accordingly, the distributing rule of different sub-audios in the wavelet domain is similar to each other, SVR can be used to learn the characteristics of them. Using the information of unmodified template positions in the low-frequency region of the wavelet domain, the SVR can be trained well. Thanks to the good learning ability of SVR, the watermark can be correctly extracted under several different attacks. The proposed watermarking method which doesn’t require the use of the original audio signal for watermark extraction can provide a good copyright protection scheme. The experimental results show the algorithm is robust to signal processing, such as lossy compression (MP3), filtering, resampling and requantizing, etc.

DC^* (Double Clustering by A^*) is an algorithm for interpretable fuzzy information granulation of data. It is mainly based on two clustering steps. The first step applies the LVQ1 algorithm to find a suitable representation of data relationships. The second clustering step is based on the A^* search strategy and is aimed at finding an optimal number of fuzzy granules that can be labeled with linguistic terms. As a result, DC* is able to linguistically describe hidden relationships among available data. In this paper we propose an extension of the DC^* algorithm, called DC $^{*} _{1.1}$ , which improves the generalization ability of the original DC^* by modifying the A^* search procedure. This variation, inspired by Support Vector Machines, results empirically effective as reported in experimental results.

This paper introduces a parallel combination of N  > 2 one class fuzzy KNN ( FKNN ) classifiers. The classifier combination consists of a new optimization procedure based on a genetic algorithm applied to FKNN ’s, that differ in the kind of similarity used. We tested the integration techniques in the case of N  = 5 similarities that have been recently introduced to face with categorical data sets. The assessment of the method has been carried out on two public data set, the Masquerading User Data (www.schonlau.net) and the badges database on the UCI Machine Learning Repository (http://www.ics.uci.edu/~mlearn/). Preliminary results show the better performance obtained by the fuzzy integration respect to the crisp one.

The term ‘missing cluster’ (MC) is introduced as an undesirable feature of fuzzy partitions. A method for detecting persistent MCs is shown to improve the choice of proper fuzzy parameter values in fuzzy C-means clustering when compared to other methods. The comparison was based on simulated data and gene expression profiles of cancer.