Optimasi Training Neural Network Menggunakan Hybrid Adaptive Mutation PSO-BP

Salnan Ratih Asriningtias, Harry Soekotjo Dachlan, Erni Yudaningtyas

Abstract


Optimization of training neural network using particle swarm optimization (PSO) and genetic algorithm (GA) is a solution backpropagation’s problem. PSO often trapped in premature convergent (convergent at local optimum) and GA takes a long time to achieve convergent and crossover makes worse the results. In this research adaptive mutation particle swarm optimization and backpropagation (AMPSO-BP) is used for training the neural network of the iris plant, breast cancer, wine, glass identification and pima indian diabetes. The addition of PSO with adaptive mutation to prevent premature convergent and BP to increase the efficiency of local searching. AMPSO-BP training results will be compared with the GA and BP. The test results showed AMPSO-BP is able to optimize the process of training the neural network. AMPSO-BP more rapidly achieve the minimum error (global minimum), fast convergent and have the ability memorization and generalization with more accurate results than the other methods.

Index Terms—Adaptive Mutation, Backpropagation, Particle Swarm Optimization, Training Neural Network.


Full Text:

PDF

Refbacks

  • There are currently no refbacks.