Learning Model Based on Electrochemical Metallization Memristor with Cluster Residual Effect

Abstract

Although a memristor model, subjected to electrochemical metallization mechanism, has been proposed based on the spontaneous decay of clusters in the previous work, it does not agree with the human forgetting accurately. Therefore, an improved model is meaningfully presented for the memristor with the cluster spontaneous decay by adding the residual effect. The former is due to the inward contraction of atoms driven by surface energy, while the latter is because of the balance of attractive and repulsive forces between atoms. The model fits well with the actual device. The forgetting is caused by the spontaneous decay. Memory retention is generated due to the added effect, which is also the internal cause of good agreement with the actual forgetting. Additionally, short‐term plasticity is converted to long‐term plasticity through the repeated learning. The efficiency of experiential learning using this model is much higher than that using the previous. It is shown that the physical mechanism of spontaneous decay in the cluster‐based channel is different from that in vacancy‐based or atom‐based channel. The model working under a non‐ideal condition with the temperature influence is discussed. Potential applications based on the model are stated.