Temperature effects modelling and compensation analysis in analogue implementation of stochastic artificial neural networks

Abstract

The implementation of Artificial Neural Networks (ANN) as CMOS analogue integrated circuits shows several attractive features. Numerous papers show that small size analogue networks operate correctly. However, real ANN applications will require large networks. On the other hand, all of the presented implementations of ANN have been supposed to be working in ideal conditions but real applications will be subject to some global perturbations. For the analogue and mixed digital/analogue implementation cases, the behaviour analysis of the neural network with perturbation conditions is thus inevitable. Unfortunately, very few papers analyse the behaviour of analogue neural network with global perturbations. In this paper we present the behaviour analysis of a CMOS analogue implementation of the synchronous Boltzmann machine model with physical temperature perturbations. The relation between the T parameter of the Boltzmann machine's model and the physical temperature of circuit has been established. Simulation results have been given, temperature effects compensation have been discussed and finally, experimental results have been presented.