Construction Of Neuromorphic Circuit Based On Memristors And Realization Of Associative Memory

Memristor is an emerging nanoelectronic component.The resistance variability and nonvolatility of the memristor are highly similar to the characteristics of biological synapses,which provides a basis for implementing neuromorphic computing in the form of hardware.Associative memory is an important manifestation of human brain intelligence and an important way for humans to recognize the world.In this dissertation,memristors are used to simulate biological synapses,and neuromorphic circuits are constructed to realize associative memory.The main contents are as follows:Aiming at the phenomenon that associative memory occurs when food and ring signals do not appear simultaneously in Pavlov associative memory,the concept of time delayed learning is proposed.A fully function memrisor-based Pavlov associative memory neuromorphic circuit with the ability of time delay learning is constructed.The constructed circuit uses the voltage control module and the time delay module to operate the synapse module,so as to realize associative memory.The circuit has functions such as learning,forgetting,variable rate learning and forgetting,time delayed learning and forgetting,and natural forgetting.The designed neuromorphic circuit based on memristors can effectively reproduce the conditioned reflex behavior of intelligent organisms,which provide a method for constructing more complex brain-like circuits.Aiming at the problem that the Hopfield neural network simulates biological synapses with resistance and cannot be repeatedly trained,a Hopfield neuromorphic circuit based on memristive synapses is constructed.The constructed circuit takes the information processing module composed of the memristive array as the core,and executes the operation under the control of the designed iterator module,thereby realizing associative memory.Synaptic weights can be repeatedly adjusted and trained to remember new patterns.The memristive Hopfield associative memory neuromorphic circuit can infer the final result from the incomplete prompt information,which provides a means for the development of computing tools that can simulate human memory.Aiming at the problem that a specific memristive array structure can only achieve a specific computing task,a memristive neuromorphic circuit with variable structure is designed.Through the fusion of the memristor array and MOSFET(Metal-Oxide-Semiconductor Field-Effect Transistor),the constructed circuit closes the unused signal input port to resist noise interference,and blocks the unused memristor from receiving the working voltage to reduce the energy consumption.The memristive neuromorphic circuit with variable structure can flexibly configure the circuit structure according to the task requirements,which provides ideas for building a general brain-like circuit that can be compatible with multiple structures.In the dissertation,with memristor as the basic component,Pavlov associative memory neuromorphic circuit,Hopfield associative memory neuromorphic circuit,and neuromorphic circuit with variable structure are constructed.The associative memory function is realized by hardware,which provides a reference for theoretical research and circuit implementation in the field of brain-like intelligence.