The use of the electrical simulator SPICE for behavioral simulation of artificial neural networks

Artificial Neural Networks (ANNs) are an alternative solution to traditional architectures for massive computation i n an increasing number of applications. Behavioral simulation of these computing paradigms requires the use of specific C A D tools to verify the correctness of structures and algorithms before starting the subsequent phases of design and implementation b y means of digital, analog or mized integrated circuits. This paper presents an approach to behavioral simulation which is independent from the technological solution used to implement the A N N . Our proposal is based upon the use of the well-known electrical simulator SPICE [8]: mapping of the A N N entities onto electrical components is described and ezperimental results are discussed.