Recurrent Neural Networks Design by Means of Multi-Objective Genetic Algorithm
Evolutionary algorithms are considered more efficient for optimal system design because they can provide higher opportunity for obtaining the global optimal solution. This paper introduces a method for construct and train Recurrent Neural Networks (RNN) by means of Multi-Objective Genetic Algorithms (MOGA). The use of a multi-objective evolutionary algorithm allows the definition of many objectives in a natural way. The case study of the proposed model is the phoneme recognition. We have shown that the proposed model is able to achieve good results in recognition tasks.
Keywords: Recurrent neural network, Genetic algorithm, Phonemes recognition, Multi-objective optimization
Download Full-Text
