On the analysis and design of neural nets with an implementation via CMOS VLSI

The author presents a mathematical framework for the analysis and consequently the design of neural networks which would guarantee the convergence of solutions and the existence of finite number of steady-state (equilibrium) solutions only. He also introduces certain conditions on the desired memories so that they can be stored and presents a synthesis for the interconnection matrix to store a given set of memories. A CMOS VLSI implementation is briefly described.<>