Design of Superior Parameterized Quantum Circuits for Quantum Image Classification

Abstract

In this paper, we present a novel quantum neural network (QNN) algorithm enhanced with transfer learning to perform multi-class image classification. The proposed QNN extracts quantum image encoding measurements through the quantum state tomography framework and passes the sampled features through the classical neural network architecture to the proposed learnable parameterized quantum circuit (PQC) followed by gradient update via quantum backpropagation. We benchmark three different PQCs to demonstrate that our proposed algorithm outperforms similar classical CNN architecture in test accuracy on CIFAR10 and MNIST datasets. Present results more prominently establish the success of PQC designs which will be further used in the design of 2D quantum convolutional neural network (QCNN).