Memristor crossbar-based Pavlov associative memory network for dynamic information correlation

Abstract

Pavlov associative memory is an important learning style of biologies to correlate input stimuli. Memristor crossbar-based circuits occupy the low manufacture burden. Based on the circuit structure of memristor crossbar array and the Pavlov associative memory mechanism, this paper proposes a memristor crossbar-based Pavlov associative memory network to realize dynamic information correlation. First, a novel synaptic circuit which consists of four MOS transistors and one memristor is proposed to act as the cell in the memristor crossbar array. Then, the memristor crossbar-based neuron circuit is designed for the construction of the memristor crossbar-based Pavlov associative memory network. Relying on associative learning, the correlations between the input stimuli are formed to store correlation information values in the memristor crossbar-based Pavlov associative memory network. Next, the stored correlation information can be read out when inputting the single stimulus. Additionally, the memristor crossbar-based Pavlov associative memory network has the forgetting process along with time to update the stored correlation information dynamically. A case study of the proposed Pavlov associative memory network circuit is implemented to realize fruit correlation information store, which proves that the proposed circuit can accomplish information correlation effectively. The proposed Pavlov associative memory network circuit has the potential to be applied in the human–computer interaction system.