In this paper, a novel heuristic structure optimization technique is proposed for Neural Network using Adaptive PSO & GA on Boolean identities to improve the performance of Artificial Neural Network (ANN). The selection of the optimal number of hidden layers and nodes has a significant impact on the performance of a neural network, is decided in an adhoc manner. The optimization of architecture and weights of neural network is a complex task. In this regard the use of evolutionary techniques based on Adaptive Particle Swarm Optimization (APSO) & Adaptive Genetic Algorithm (AGA) is used for selecting an optimal number of hidden layers and nodes of the neural controller, for better performance and low training errors through Boolean identities. The hidden nodes are adapted through the generation until they reach the optimal number. The Boolean operators such as AND, OR, XOR have been used for performance analysis of this technique.
Sahu, Amaresh; Panigrahi, Sushanta; and Pattnaik, Sabyasachi
"Optimization of ANN Structure Using Adaptive PSO & GA and Performance Analysis Based on Boolean Identities,"
International Journal of Computer and Communication Technology: Vol. 4:
4, Article 7.
Available at: https://www.interscience.in/ijcct/vol4/iss4/7