Study of Principle Component Analysis and Learning Vector Quantization Genetic Neural Networks

Abstract

In this work, the Genetic Algorithm (GA) is used to improve the performance ofLearning Vector Quantization Neural Network (LVQ-NN), simulation results show thatthe GA algorithm works well in pattern recognition field and it converges much fasterthan conventional competitive algorithm. Signature recognition system using LVQ-NNtrained with the competitive algorithm or genetic algorithm is proposed. This schemeutilizes invariant moments adopted for extracting feature vectors as a preprocessing ofpatterns and a single layer neural network (LVQ-NN) for pattern classification. A verygood result has been achieved using GA in this system. Moreover, the PrincipleComponent Analysis Neural Network (PCA-NN) which its learning technique isclassified as unsupervised learning is also enhanced by hybridization with the geneticalgorithm. Three algorithms were used to train the PCA-NN. These are GeneralizedHebbian Algorithm (GHA), proposed Genetic Algorithm and proposed HybridNeural/Genetic Algorithm (HNGA).

Keywords

• neural networks
• genetic algorithms
• principle component analysis
• learning vector quantization
• signature recognition