Implementation of Artificial Neural Networks Trained by Particle Swarm Optimization using Multi-Phase Switched–Capacitor Circuits

Abstract

In this paper, a proposed design of Artificial Neural Networks Trained by Particle Swarm Optimization using multi-phase switched-capacitor circuits is presented. Swarm intelligence is based on collective behavior of self organized group of agents. Each agent follows a relatively simple set of rules and interacting with its local surrounding. Particle Swarm Optimization (PSO) has been an increasingly interesting topic in the field of computational intelligence. PSO is another optimization algorithm that falls under the soft computing address. One application of PSO has tremendous success is in the field of Artificial Neural Networks (ANNs) training. In this paper an adaption of the ANN weights using PSO is proposed as a mechanism to improve the performance of ANN. For this purpose we have modified the MATLAB PSO toolbox to be suitable with neural application. In neural networks, the multiplier is needed to deal with the learning of weights, and the generation of associated outputs therefore, a proposed design of multiplier circuit using multi-phase switched-capacitor circuit that can be implemented in CMOS technology. Generating multiple clock sources is a common requirement for the designing multi-phase switched-capacitor circuits so; a proposed design of multi-phase clock generator is presented which produces sequential non-overlapping clock pulses. The proposed design of multi-phase switched-capacitor neuron and its corresponding “synapses” also presented in details. Simulation results are presented using EWB package, which illustrates the validity of the proposed switched capacitor circuit's designs.