Volume 8, Número 3
Abstract
Download
Cover and Table of Contents, Volume 8, Issue 3, 2010 of the Learning and Nonlinear Models (L&NLM) - Journal of the Brazilian Society of Neural Networks
Lariza L. de Oliveira, Gabriela F. Persinoti, Silvana Giuliatti & Renato Tinós
Abstract
Download
In this work, Genetic Algorithm (GA) is employed in feature selection for the classification of medicinal plants with snake venom-neutralizing properties. The classification is performed using an Artificial Neural Network (ANN), which indicates the medicinal plants with anti-snake venom action as output when an amino acid sequence of snake venom is presented in its input. GAs and ANNs are Artificial Intelligence techniques and have been used in several similar optimization and classification problems. Here, the feature selection system is implemented using the classification error rate of the training set and the number of attributes as the fitness of each individual of the GA. The validation results for the classification system indicate that ANNs can be used to aid the selection of medicinal plants with snake venom-neutralizing properties. Also, feature selection based on GAs can help researches to select amino acids sequences of the snake venoms which can be important to the interaction with medicinal plants compounds.
Vinicius Tragante do Ó & Renato Tinós
Abstract
Download
In the Genetic Algorithm (GA) with the standard random immigrants approach, a fixed number of individuals of the current population are replaced by random individuals in every generation. The random immigrants inserted in every generation maintain, or increase, the diversity of the population, what is advantageous to GAs applied to complex problems like the protein structure prediction problem. The rate of replaced individuals in the standard random immigrants approach is defined a priori, and has a major influence on the performance of the algorithm. In this paper, we propose a new strategy to control the number of random immigrants in GAs applied to the protein structure prediction problem. Instead of using a fixed number of immigrants per generation, the proposed approach controls the number of new individuals to be inserted in the generation according to a self-organizing process. Experimental results indicate that the performance of the proposed algorithm in the protein structure prediction problem is superior or similar to the performance of the standard random immigrants approach with the best rate of individual replacement.
Igor S. Peretta, Gerson F. M. Lima, Josimeire A. Tavares & Keiji Yamanaka
Abstract
Download
The use of voice commands as a new way of interaction between man and machine is the subject of several researches in recent years and has already been produced commercial and freeware applications. However, considering the achieved results, there is still a great development potential in this area, particularly in Brazilian Portuguese language. This work proposes: 1. an efficient method of detecting spoken word boundaries from a recorded signal, using Teager Energy Operator and FIR Filter; 2. the use of wavelet transform and wavelet packet filter bank as a main tool for feature extraction to feed a multi-layer artificial neural network to recognize a limited vocabulary of voice commands. The system was developed using a dataset of spoken words from 50 speakers, using normal pronunciation speed and in an environment without any noise control. Tests with the system show a very good classification rate and noise robustness.
Ricardo de A. Araújo
Abstract
Download
This work presents a new evolutionary morphological-rank-linear approach in order to adjust time phase distortions in financial time series forecasting, overcoming the random walk dilemma. The proposed approach, referred to as Evolutionary Morphological-Rank-Linear Forecasting (EMRLF) method, consists of an intelligent hybrid model composed of a Morphological-Rank-Linear (MRL) filter combined with a Modified Genetic Algorithm (MGA), which performs an evolutionary search for the minimum number of relevant time lags capable of a fine tuned characterization of the time series, as well as for the initial (sub-optimal) parameters of the MRL filter. Each individual of the MGA population is improved using the Least Mean Squares (LMS) algorithm to further adjust the parameters of the MRL filter, supplied by the MGA. After built the prediction model, the proposed method performs a behavioral statistical test with a phase fix procedure to adjust time phase distortions that can appear in the modeling of financial time series. An experimental analysis is conducted with the proposed method using two real world stock market time series according to a group of performance metrics and the results are compared to both MultiLayer Perceptron (MLP) networks and a more advanced, previously introduced, Time-delay Added Evolutionary Forecasting (TAEF) method.
S. Ribas, M. H. P. Perché, I. M. Coelho, P. L. A. Munhoz, M. J. F. Souza & A. L. L. Aquino
Abstract
Download
This paper introduces MaPI, a framework that implements the MapReduce abstraction. Using MaPI, the user is
able to implement parallel applications without worrying about the messages transmission between the processes, or how the
system will do the parallelization. Furthermore, all the implementation made by the user can be sequential. In order to demon-
strate how the framework works, it is used to parallelize an optimization heuristic algorithm applied to a classical optimization
problem, the Traveling Salesman Problem. The results show the efficiency of the framework as a tool to help the development of
parallel optimization procedures.
