A FEEDFORWARD ARTIFICIAL NEURAL-NETWORK-BASED ON QUANTUM EFFECT VECTOR-MATRIX MULTIPLIERS

The vector-matrix multiplier is the engine of many artificial neural n etwork implementations because it can simulate the way in which neuron s collect weighted input signals from a dendritic arbor. We present a new technology for building analog weighting elements that is theoreti cally capable of densities and speeds far beyond anything that convent ional VLSI in silicon could ever offer. To illustrate the feasibility of such a technology, we have built a small three-layer feedforward pr ototype network with five binary neurons and six tri-state synapses an d used it to perform all of the fundamental logic functions: XOR, AND, OR, and NOT.