Memristor-Based Directional Forgetting Neural Network Circuit With Emotion Memory and Its Application in Intelligent Robots

Most neural networks based on memristor only consider the relationship between emotion and memory, but ignore the relationship between emotion and directional forgetting. This article presents a neural network circuit based on memristors, which can achieve targeted forgetting with self-correlation and emotional memory. The designed circuit is mainly composed of memory module, self-correlation module, amygdala module, emotion module, and association neuron module. Memory module can dynamically adjust the time for memory formation and the time for memory decay according to the requirements of the task. Association between learning materials and the characteristics of the individual can be achieved through the self-correlation module. The learning material signals are transformed into positive learning material signal and negative learning material signal through the amygdala module. Emotion module and the association neuron module are used to determine whether the emotionality of the learning materials is consistent with one's own emotionality. The feasibility of the function of the proposed circuit is verified through Simulation Program with integrated circuit emphasis. This circuit provides more references for the further development of brain-like intelligence.