BACTERIAL FORAGING OPTIMIZATION ALGORITHM FOR NEURAL NETWORK LEARNING ENHANCEMENT

Abstract

Backpropagation algorithm is used to solve many real world problems using the concept of  Multilayer Perceptron. However, the main disadvantages of Backpropagation are its convergence rate is relatively slow, and it is often trapped at the local minima. To solve this problem, in literatures, evolutionary algorithms such as the Particle Swarm Optimization algorithm has been applied in feedforward neural network to optimize the learning process in terms of convergence rate and classification accuracy but this process needs longer training time. To provide alternative solutions, in this study, Bacterial Foraging Optimization Algorithm has been selected and applied in feedforward neural network to enhance the learning process in terms of convergence rate and classification accuracy. One of the main processes in Bacterial Foraging Optimization algorithm is the chemotactic movement of a virtual bacterium that makes a trial solution of the optimization problem. This process of chemotactic movement is guided to make the learning process of Artificial Neural Network faster.  The developed Bacterial Foraging Optimization Algorithm Feedforward Neural Network is compared against Particle Swarm Optimization Feedforward Neural Network (PSONN). The results show that BFOANN gave a better performance in terms of convergence rate and classification accuracy compared to PSONN.