Electronic neural circuits with simulated annealing

Summary form only given, as follows. A promising approach to implementing electronic neural networks is to design and fabricate special-purpose VLSI chips, because the software execution of simulated annealing on conventional computers is very slow. A hardware realization of simulated annealing is described using the analog between the temperature in the Boltzmann machine and the amplifier gain in the electronic neural circuits. Detailed requirements for the annealing schedule are analyzed. Experimental results on the transfer characteristics of 4-bit Hopfield neural-based analog-to-digital converters with the simulated annealing technique are presented.<>