A novel architecture for implementing large-scale Hopfield neural networks using CDMA communication technology

We propose a novel LSI architecture that allows silicon LSIs to implement mutual-coupled neural networks. The architecture reduces wiring areas of Hopfield neural networks by using CDMA protocols for communicating between neurons. As an example, we propose a CDMA-Hopfield neural network, aiming at examining retrieval properties of the network. Extensive simulation results indicated that the CDMA-Hopfield neural network of N neurons could retrieve signal patterns from P memory patterns when P/N ≈ 0.1.