Design Of New Memristive Neural Network Circuits

With the rapid development of information technology,people have higher and higher requirements on electronic devices and circuit integration.The traditional transistor faces severe challenges in size and performance.Memristor is a new circuit element,it has low power consumption,nanoscale,memory characteristics and so on.It has broad potential application prospects in nonvolatile memory,digital logic circuits and neural networks.The memristor's nanoscale and memory properties make it the best choice for constructing a new generation of artificial synapses.On the basis of introducing memristor and neural network theory,this paper designs memristor synapse and new memristor neural network model based on HP nonlinear drift model,and realizes the high biological activities of learning,associative memory and forgetting process.The main contents of this paper are as follows:(1)In order to determine whether a memristor or a memcapacitor is more suitable for simulation of synapses,we propose two memcapacitor models,the memcapacitance calculation formulas were deduced,and compared with two memristor models,the analysis results show that the memristor is more suitable for the design of the synapses,and the selected window function of HP memristor model is used in the design of neural network.(2)Based on the HP memristor model,the AND gate OR gate logic circuit is designed,the working principle of the gate circuit is analyzed,the logic gate circuit is simulated by SPICE software,and its logic relationship is verified,which lays a foundation for the design of neural network circuit in this paper.(3)An improved CMOS neuron circuit is designed and its working principle is theoretically analyzed and simulated;This paper proposes a new modification rule of synaptic weight,which not only learning and two kinds of forgetting,but also the modification rule of the third kind of forgetting,which can well explain the changes of synaptic weight and lay a theoretical foundation for the subsequent design and application of associative memory neural network.(4)Based on the above memristor model,memristor logic gate circuit,CMOS neuron circuit and synaptic weight modification rules,a new memristor neural network circuit is designed to realize biological activities including learning,associative memory and forgetting.The principle of the neural network is analyzed theoretically,and the function of the memristor neural network is simulated by SPICE software.