Research On Quantum Neural Network Model Based On Josephson Junction

The advance of Josephson junction neuron provides a new tool to study the neural network.Combined with the Josephson junction neuron and quantum computing, this paper researchesquantum neural network and its performance based on Josephson junction.Firstly, this paper studies biological neuron characteristics of the Josephson junction, such asaction potential, refractory period, threshold, binary and so on. The simulation results verify theJosephson junction neuron has dynamic properties of biological neuron.Secondly, the Josephson junction quantum neuron and its characteristic is researched. On thebasis of the characteristics of the Josephson junction, a new quantum neuron model based onJosephson junction is proposed, its learning algorithm is designed and its convergence propertiesare analyzed. Combined with quantum computation theory, the quantum logic functions ofJosephson junction quantum neuron are analyzed, the analysis shows that the Josephson junctionquantum neuron can realize quantum not-gate, Hadamard gate and XOR operation which can’t beachieved by a single classical neuron, and has the same nonlinear mapping ability as the two-layerclassical neural network.Finally, the quantum neural network model and its characteristics are studied. The neuralnetwork based on Josephson junction and its learning algorithm is designed, this model is applied tofunction approximation. The experimental results show that the performance of the Josephsonjunction neural network is better than the neural network with sigmoid function. On this base, thispaper designs Josephson junction activation function based on quantum level concept, proposesquantum neural network model based on quantum level Josephson junction, designs the learningalgorithm of this model. This model is applied to three classification problems. The simulationresults show that the quantum neural network is more accurate than Josephson junction neuralnetwork.