Catálogo de publicaciones - libros

The system is designed to enhance the specific features of each user, without increasing the differences between users in what concerns the level of under-standing or the ability to creatively use the acquired knowledge.

The paper describes also a framework for building new courses or updating existing ones by choosing learning objects developed at universities that are members a federated learning system. The aim of is to assisting a lecturer in collecting learning objects closest to the lecturer’s vision on what a subject should contain and how the content should be presented, and to present a student with contents, tailored according to student’s individual learning preferences.

An adaptive tutoring system for an Operating System course is presented. The architecture, based on sequencing graphs, that supports an adaptive sequencing of the learning units is described. The content structure is presented as well. The system is now in use in regular university courses and results of this experience will be published in the future.

In this paper we present a novel system that utilizes molecular-level information for medical diagnosis. It accepts high dimensional vectors of gene expressions, quantified by means of microarray image analysis, as input. The proposed system incorporates various data pre-processing methods, such as missing values estimation and data normalization. A novel approach to the classification of gene expression vectors in multiple classes that embodies vari-ous gene selection methods has been adopted for diagnostic purposes. The pro-posed system has been extensively tested on various, publicly available data-sets. We demonstrate its performance for prostate cancer diagnosis and corn-pare its performance with a well established multiclass classification scheme. The results show that the proposed system could be proved a valuable diagnostic aid in medicine.

This paper presents a synergistic methodology for automatically recording, monitoring and interrelating changes occurred in invivo bio-cells without any user’s assistance. The methodology presented here combines several techniques, such as projection functions, registration, segmentation with region synthesis, local-global graphs and stochastic Petri-nets. Each of these techniques produces complementary results and the synergistic combination of them generates a methodology that produces the bio-signatures of bio-cells in sequences of images. Illustrative results are also provided.

The aim of this work is to evaluate a robotic system for remote performance of minimally invasive procedures with real-time magnetic resonance imaging (MRI) guidance inside clinical cylindrical scanners. In these studies, the operator had no physical access to the subject and used MR images and video from the observation camera in the scanner to control the robot. The control software allowed manual and semi-automated control modes and included components for collision avoidance, with the subject or the gantry of the scanner, and on-the-fly adjustment of the MR imagine plane to visualize the procedure. Studies were performed initially on phantoms and lastly on a pig inside a standard clinical cylindrical 1.5 Tesla MR scanner.

The selection of gene subsets that retain high predictive accuracy for certain cell-type classification, poses a central problem in microarray data analysis. The application and combination of various computational intelligence methods holds a great promise for automated feature selection and classification. In this paper, we present a new approach based on evolutionary algorithms that addresses the problem of very high dimensionality of the data, by automatically selecting subsets of the most informative genes. The evolutionary algorithm is driven by a neural network classifier. Extensive experiments indicate that the proposed approach is both effective and reliable.

Cancer classification is one major application of microarray data analysis. Due to the ultra high dimension of gene expression data, efficient feature selection methods are in great needs for selecting a small number of informative genes. In this paper, we propose a novel feature selection method based on mutual information and rough set (MIRS). First, we select some top-ranked features which have higher mutual information with the target class to predict. Then rough set theory is applied to remove the redundancy among these selected genes. Binary particle swarm optimization (BPSO) is first proposed for attribute reduction in rough set. Finally, the effectiveness of the proposed method is evaluated by the classification accuracy of SVM classifier. Experi-ment results show that MIRS is superior to some other classical feature selec-tion methods and can get higher prediction accuracy with small number of fea-tures. Generally, the results are highly promising.

In recent years, computational diagnostic tools and artificial intelligence techniques provide automated procedures for objective judgments by making use of quantitative measures and machine learning. The paper presents a Support Vector Machine (SVM) approach for the prognosis and diagnosis of breast cancer implemented on the Wisconsin Diagnostic Breast Cancer (WDBC) and the Wisconsin Prognostic Breast Cancer (WPBC) datasets found in literature. The SVM algorithm performs excellently in both problems for the case study datasets, exhibiting high accuracy, sensitivity and specificity indices.

Source code author identification deals with the task of identifying the most likely author of a computer program, given a set of predefined author candidates. This is usually. based on the analysis of other program samples of undisputed authorship by the same programmer. There are several cases where the application of such a method could be of a major benefit, such as authorship disputes, proof of authorship in court, tracing the source of code left in the system after a cyber attack, etc. We present a new approach, called the SCAP (Source Code Author Profiles) approach, based on byte-level n-gram profiles in order to represent a source code author’s style. Experiments on data sets of different programming language (Java or C++) and varying difficulty (6 to 30 candidate authors) demonstrate the effectiveness of the proposed approach. A comparison with a previous source code authorship identification study based on more complicated information shows that the SCAP approach is language independent and that n-gram author profiles are better able to capture the idiosyncrasies of the source code authors. Moreover the SCAP approach is able to deal surprisingly well with cases where only a limited amount of very short programs per programmer is available for training. It is also demonstrated that the effectiveness of the proposed model is not affected by the absence of comments in the source code, a condition usually met in cyber-crime cases.

Agent-building tools have an important role in popularizing and application of agent technology. This paper describes a new agent-programming tool AJA. AJA consists of two programming languages: HADL for defining of higher-level agent constructs and Java+ for low-level programming of these constructs. Among other interesting features AJA presents an original approach of incorporating artificial neural nets, into a programming language.