Memristor-Based Neural Network Circuit of Duple-Reward and Duple-Punishment Operant Conditioning With Time Delay
Junwei Sun, Yuanpeng Xu, Peng Liu, Yanfeng Wang
Abstract
Currently, the research in memristor-based associative memory neural networks pays more attention to classical conditioning and lays less attention to operant conditioning. Moreover, a single reinforcing stimulus is applied in the most studies of operant conditioning. In this paper, a memristor-based duple-stimulus operant conditioning circuit is designed. The circuit utilizes duple stimuli which include two stimuli and realizes the effect of stimuli with two intensities on operant conditioning. In addition, the circuit is applied to a memristor-based neural network circuit of duple-reward and duple-punishment operant conditioning with time delay. The application circuit realizes duple-reward and duple-punishment operant conditioning with time delay on the basis of duple-stimulus operant conditioning circuit. Meanwhile, the factors that affect operant conditioning such as satiety and the immediacy of stimuli are implemented in the circuit. Different delays for duple-reward module and duple-punishment module are designed. The application circuit is closer to the practical application and is more consistent with biological characteristics. It provides more references for neural networks of operant conditioning with further development.