Volume 8, Número 2
Abstract
Download
Cover and Table of Contents, Volume 8, Issue 2, 2010 of the Learning and Nonlinear Models (L&NLM) - Journal of the Brazilian Society of Neural Networks
Carlos Alexandre R. Fernandes, João Cesar M. Mota & Gérad Favier
Abstract
Download
Multiple-input multiple-output (MIMO) Volterra models have applications in different areas, including telecommunications. An overview of the modeling of nonlinear communication channels using MIMO Volterra models is presented in this paper. First, the development of an equivalent baseband discrete-time representation of a single-input single-output (SISO) Volterra system is carried out. This development constitutes the basis for several versions of discrete-time equivalent baseband MIMO Volterra systems presented in the sequel. The spectral broadening provided by a Volterra system on the equivalent baseband received signals is shown by calculating the frequency domain representation of the Volterra channel output. Some important block structured nonlinear MIMO models are also described, with their link to MIMO Volterra models. Finally, some applications of such models for communication systems are briefly discussed.
Vitor H. Ferreira, André Lazzaretti, Hugo V. Neto, Rodrigo Riella & Julio Omori
Abstract
Download
This work proposes a method for automatic classification of oscillographies corresponding to faults and events related to service quality in electricity distribution networks. The proposed method is divided in two stages: pre-processing and classification of events. In the first stage, oscillography signals are decomposed using the wavelet transform. The energy present in each sub-band of the wavelet domain is then computed in order to compose feature vectors, which are fed to the automatic classifiers of the second stage. The classifiers investigated are based on Multi-Layer Perceptron (MLP) feed-forward artificial neural networks and Support Vector Machines (SVM), which are able to promote feature selection and network complexity control simultaneously. Experiments using simulated data yielded promising results in service quality event classification.
Matheus de Souza A. Silva, Marcio T. Mine, Luiz S. Ochi & Marcone J. F. Souza
Abstract
Download
This paper proposes a hybrid evolutionary algorithm for getting an approximate solution for the Prize Collecting Covering Tour Problem (PCCTP). The proposed algorithm combines heuristic strategies based on Iterated Local Search, Variable Neighborhood Descent, Path Relinking and GENIUS procedures. Computational results on a set of instances illustrate the effectiveness and the robustness of the proposed heuristic.
Luiz P. Caloba, Max S. Dutra, Alvaro D. Orjuela & Ivanovich Lache
Abstract
Download
The neural network's relevance for clustering different kind of data has encouraged diverse works in order to optimize their implementation. Between the areas of research, exist ones that seek to reduce the number of neurons needed to define a specific class. In this work the authors present an algorithm that allows the reduction of number of neurons width variable radius in ART's networks, the procedure used is listed as semi-supervised, but has features that allow its application in conjunction with not supervised techniques. Several tests were performed to assess the effectiveness of this method, two of the most important and informative examples are presented.
Tiago M. da Silva, Antônio de P. Braga & Wilian S. Lacerda
Abstract
Download
This work presents an on-chip learning of artificial neural
networks in a FPGA multiprocessor system, where each neuron
is implemented in a soft-core processor. In order to take maximum
advantage of the distributed architecture, a pipelined version of
the on-line back-propagation algorithm is used, providing a high
degree of parallelism between neuron layers and, hence, a higher
speed-up in relation to a sequential implementation.
