MNOS/CCD circuits for neural network implementations

A description is given of a circuit technology which particularly suits the characteristics of neural networks, especially those with the emphasis on adaptation and learning. The authors describe the basic devices and circuit elements. They discuss the computational operations that these circuits can perform. A description is given of a Hopfield neural network integrated circuit that the authors designed, fabricated, and tested. They believe that this chip is the first integrated circuit neural network to demonstrate complete learning on the chip.<<ETX>>