Catálogo de publicaciones - libros

Traditional gene selection methods often select the top–ranked genes according to their individual discriminative power. We propose to apply feature evaluation measure broadly used in the machine learning field and not so popular in the DNA microarray field. Besides, the application of sequential gene subset selection approaches is included. In our study, we propose some well-known criteria (filters and wrappers) to rank attributes, and a greedy search procedure combined with three subset evaluation measures. Two completely different machine learning classifiers are applied to perform the class prediction. The comparison is performed on two well–known DNA microarray data sets. We notice that most of the top-ranked genes appear in the list of relevant–informative genes detected by previous studies over these data sets.

Handheld devices, which have been widely adopted to clinical environments for medical data archiving, have revolutionized error-prone manual processes of the past. Meanwhile, the use of devices has been reported to be limited to data entries and archiving, without fully leveraging their computing and retrieval capabilities. This paper studies a hybrid system which complements current state-of-art, by combining intelligent retrieval techniques developed over middleware environments for and flash-aware data management techniques for . By enabling intelligent ranked retrieval, the limited resources of handheld devices, eg., limited display and computation capabilities, can be utilized effectively, by selectively retrieving the few most relevant results. However, to achieve this goal, we need a hybrid approach, bridging middleware-based ranked retrieval techniques to optimize for flash memory storage, as typically adopted by handheld devices. We address this newly emerging challenge and propose a flash-aware framework , which we empirically validate its effectiveness over baseline alternatives.

This paper describes a hybrid intelligent medical tutor for atheromatosis. The tutor is called INTATU (INTelligent Atheromatosis TUtor). INTATU provides adaptive tutoring on Atheromatosis to various classes of users depending on their interests, background medical knowledge and computer skills. The adaptivity results from user modelling that is based on stereotypical knowledge about the potential users (patients, patients’ relatives, doctors, medical students, etc.). The inference mechanism uses a hybrid combination of rule-based reasoning of double stereotypes and decision making techniques.

In data mining, hybrid intelligent systems present a synergistic combination of multiple approaches to develop the next generation of intelligent systems. Our paper presents an integration of a Combined Multiple Models (CMM) technique with an evolutionary approach that is used for tuning of parameters. Proposed hybrid classifier was tested in microarray analysis domain. This domain was chosen intentionally, because of the nature of Combined Multiple Models classifiers that are specialized in solving problems with high dimensionality and contain low number of samples. Evolutionary tuning of parameters in combination with validation dataset enables fine tuning of parameters that are usually set to pre-defined values. Using this technique we made another step in leveling the accuracy of comprehensible classifiers to those represented by ensembles of classifiers.

Accurate prediction of protein structures is very important for many applications such as drug discovery and biotechnology. Building side chains is an essential to get any reliable prediction of the protein structure for any given a protein main chain conformation. Most of the methods that predict side chain conformations use statistically generated data from known protein structures. It is a computationally intractable problem to search suitable side chains from all possible rotamers simultaneously using information of known protein structures. Reducing the number of possibility is a main issue to predict side chain conformation. This paper proposes an enumeration based similarity search algorithm to predict side chain conformations. By introducing “beam search” technique, a significant number of unrelated side chain rotamers can easily be eliminated. As a result, we can search for suitable residue side chains from all possible side chain conformations.

In this paper, we present the design, implementation and evaluation of HIGAS, a hybrid intelligent system that deals with diagnosis and treatment consultation of acid-base disturbances based on blood gas analysis data. The system mainly consists of a fuzzy expert system that incorporates an evolutionary algorithm in an off-line mode. The diagnosis process, the input variables and their values were modeled based on expert’s knowledge and existing literature. The fuzzy rules are organized in groups to be able to simulate the diagnosis process. Differential evolution algorithm is used to fine-tune the membership functions of the fuzzy variables. Medium scale experimental results show that HIGAS does better than its non-hybrid version, non-experts and other previous computer-based approaches.