A Fully-Connected Micro-extended Analog Computers Array Optimized by Particle Swarm Optimizer

摘要

The micro-Extended Analog Computer(uEAC) is a novel hardware implementation of Rubel's EAC model. In this study, we first analyse the basic uEAC mathematical model and two uEAC extensions with minus-feedback and multiplication-feedback, respectively. Then a fully-connected uEACs array is proposed to enhance the computational capability, and to get an optimal uEACs array structure for specific problems, a comprehensive optimization strategy based on Particle Swarm Optimizer(PSO) is designed. We apply the proposed uEACs array to Iris pattern classification database, the simulation results verify that all the uEACs array parameters can be optimized simultaneously, and the classification accuracy is relatively high.