ANN Circuit Application of Complementary Resistive Switches

Abstract

Artificial neural networks are successfully used for classification, prediction, estimation, modeling and system control. However, artificial neural networks integrated circuits are expensive and not matured enough. Memristors or memristive systems which show a nonvolatile memory behavior has a high potential for use in artificial neural network circuit applications. Some memristive synapse or memristive neural network applications already exist in literature. The complementary memristor or resistive switch memories have been suggested as an alternative to one-cell memristor memories. Their sensing is more difficult and complex than the others. The complementary memristor memory topologies with a sensing node are also inspected in literature. To the best of our knowledge, a neural network circuit which is based on the complementary resistive switches with a sensing/writing node does not exist in literature yet. In this paper, several neural network circuits which are based on the complementary resistive switches with a sensing/writing node have been designed and examined for the first time in literature. Their analysis are given and simulations are performed to verify their operation. We expect that such a complementary resistive switch implementation may find use in artificial neural networks chips in the future.