Research On Memristor Synapse And Its Implementation In Artificial Neural Network

Because of high parallelism of its architecture and event-driven pattern of computing in neural network,the human brain can perform better in some complex tasks with less space and time cost as well as energy consumption compared with the traditional computer.In order to mimic the parallel feature of human brain,how to fabricate,optimize,and integrate electronic devices which can emulate the function of biological neurons and synapses becomes an important research topic in recent years.The main content of this essay is to discuss and analyze the fabrication process of memristor artificial synapses and the approach to improve its indicators when it is implemented in artificial neural networks.Firstly,this paper introduces the background information and research significance of artificial synapses.By illustrating the basic characteristics of biological synapses,this paper will discuss about the desired merits of its artificial counterparts and stress on the recent progress on phase change memory,memristor and novel field-effect transistor which are served as artificial synapse.The needed experimental facilities and process will also be introduced by elucidating the fabrication process of memristors.The characteristics exhibited in the current measurement will lead to the discussion of memristors with two different resistance switching mechanisms and their similarities and differences.Furthermore,the drawbacks witnessed in the current measurement will be improved by proposal of a new approach.The experimental results prove the validity and feasibility of our proposed method.Finally,this paper will compare the influences of different non-ideal characteristics in the performance of artificial neural network,especially recognition accuracy and learning rate,in a face classification task,which further demonstrate the availability of proposed approach.