Circuit Design Of Memristive Hopfield Neural Network And Its Applications

The development of traditional artificial neural networks has solved many complicated problems.The hardware implementation of neural network algorithms has always been a hotspot and a difficult task for researchers.In traditional neural network circuits,resistors or CMOS devices are used as synapses.Once the weight of the synapse is fixed,it can not be modified.As a result,the development of neural network hardware has been limited and can not handle various problems flexibly.Memristor is an emerging simple two-terminal component with nanometer size,low power consumption,non-volatile and variable resistance.It can replace resistor or CMOS device as synapse in neural network.The research goal of this thesis is to design a memristive Hopfield neural network circuit through which image restoration and sparse coding are applied.In this thesis,the memristor is combined with the traditional Hopfield neural network,and a memristive Hopfield neural network circuit is designed.Using memristor as the synapse of the neural network,the flexibility of the whole circuit is improved,and the energy function is introduced to analyze the stability of this design.The image restoration problem can be transform to the error function minimization problem with constraints.The Hopfield neural network and energy function are used to transform the image restoration problem into an optimization problem,and the image restoration problem is solved by the designed memristive Hopfield neural network circuit.The purpose of the sparse coding algorithm is to find a set of basis vectors that enable the input to be represented as a linear combination of these base vectors.The sparse coding problem can also be simplified as minimize problems,and the memristive Hopfield neural network circuit also successfully solves the sparse coding problems.The research results of this thesis are mainly to design the memristive Hopfield neural network circuit,which can realize the function of optimal solution and realize the application of image restoration and sparse coding.Compared with the software implementation of the algorithm,the processing speed of the hardware circuit is greatly improved,and can be applied to artificial intelligence chips in the future.