Research On Neuron Circuit With The Property Of Nonassociative Learning

Neuron circuit is an important part of neuromorphic computing system.Compared with traditional neuron circuits based on CMOS devices,neuron circuits based on nano-scale memristors and simple electronic devices have the advantages of low operation power consumption and high integration density.These neuron circuits based on nano-scale memristors and simple electronic devices have important application value in neuromorphic computing system.In this paper,we introduced nonassociative learning principles into the design of neuron circuit based on memristors,which makes the response to unimportant repetitive stimuli decreased and the response to important strong stimuli increased.With this"pre-screening"of input stimuli,the neuron circuit based on memristors can be more efficient.The second chapter summarized the related problems of nonassociative learning neuron circuits.The biological principle of neuron circuit and memristor models used in neuron circuit design are expounded.Two kinds of neuron circuit models are compared in this chapter.The characteristics of nonassociative learning and the research status of nonassociative learning neuron circuit are also introduced.In the third chapter,we analyzed the characteristics of memristors used in nonassociative learning neuron circuit.A multi-valued memristor and a threshold-switched memristor are fabricated.And we established the memristor models based on the result of testing.Section 3.1 focused on the characteristics of memristors used in the design of nonassociative learning neuron circuit.In Section 3.2,according to the requirements of circuit design,we fabricated a Pd/HfO₂/WO_x/W multi-level memristor and an Ag/SiTe/TiN threshold switching memristor and tested their electrical characteristics.And we established the models of two kinds of memristors based on its electrical characteristics.In the fourth chapter,we designed a nonassociative learning neuron circuit based on the models we established and simulated its functions.In section 4.1,we introduced the biological principle of nonassociative learning.The 4.2 section introduced the design of the nonassociative learning neuron circuit.Section 4.3 simulated the function of the circuit.The simulation results show that the circuit can effectively simulate the habitual and sensitive nonassociative learning and improve the performance of neuron circuit.