Numerical simulations of quantum neural network and quantum circuit

Abstract

A ground-breaking machine learning, quantum neural network, has been proposed in the last few years. Its operation consists of the elements of classical machine learning and quantum computing. Evidence shows its computational power and speed is much higher than the power and rate of classical artificial neural network. However, many properties in the quantum neural network are still unknown or can not be understood. This study would enhance our understanding of the principle of quantum neural networks and related quantum phenomena. Key concepts of classical machine learning and quantum machine learning were described. The algorithm of quantum neural networks and classical artificial neural networks was briefly explained. In this work, simulations on quantum circuits have been executed. A Hadamard gate and two CNOT gates have operated on a quantum circuit. The circuit has reached an EHZ state. It revealed the simulation could imitate an actual setting and generate the right results. It introduced a few ways to describe outcomes and operations in the programming environment. It showed how to make measurements on qubits in the simulation. The constructed circuit in the simulations could output values that matched the theory. A QNN and a two-layer QNN were customed and performed forward and backward passes. They could calculate answers in both processes but there was no physical meaning for their answer and re-uploading functions for those QNN. The lack of interpretation of data encoding phrase and measurement phrase in the simulation and missing reuploading features in QNN simulation have been suggested as two main constraints. They could be countered by developing additional data encoding and measurement phrases packages and introducing new adjusting functions for QNN.