Research On Neural Network Model And Algorithm Based On Quantum Circuits

Quantum neural network is a novel computing model that combines the quantum computation with classical artificial neural networks.In theory,it is more intelligent than classical neural network and has more effective learning and generalization capabilities.At the same time,quantum neural networks also show great advantages in saving data processing resources and increasing computing speed.A major difficulty in the field of quantum neural network research is how to combine the linear reversibility of quantum computation with the nonlinear mapping structure of classical neural networks,that is,how to realize the quantum evolution of neural networks.In this article,combining the classical artificial neural network model and the effective quantum algorithm,a quantum neuron and its network model are constructed based on quantum circuits.Specific completion of the following research content.Firstly,aiming at the difficulty of transforming classical neurons into quantum forms directly,this article constructs a quantum neuron model based on quantum circuits,which meets the nonlinear computing characteristics of classical neurons and follows the laws of quantum computation.The scheme is to encode the input and weight of the neuron into the state in computational basis separately,then apply a controlled unitary gate containing the weight of neuron to the input quantum state,and finally obtain the output of the neuron by the quantum phase estimation.And the quantum circuit that realizes the entire quantum neuron and its various functional modules is given,so that the model algorithm can theoretically run on a quantum computer.Finally,IBM Qiskit open source quantum development kit is used to carry out simulation experiments,and the quantum simulator is used to implement the model and a series of logical operations to verify its effectiveness.Secondly,aiming at the failure to build a more general quantum neural network,this article builds the constructed quantum neuron model into a corresponding quantum feed-forward neural network model,and proposes its corresponding learning algorithm.The input layer,output layer and intermediate hidden layer of the neural network model are all expressed in the form of quantum states,so that it can use the unique properties of quantum computation such as superposition and parallelism to improve efficiency in the training and learning process.Then the quantum search algorithm is introduced as the learning algorithm of the neural network model.The results of simulation experiments are used to analyze the advantages of the quantum neural network compared to the classical neural network,and finally discuss its practical application in real life.