Comparative Analysis of Quantum Encoding Methods in Quantum Neural Networks for Image Classification on the MNIST Dataset

Abstract

Quantum computing, a paradigm that utilizes the concepts of quantum mechanics, has emerged as a transformative force in various fields, including machine learning. Within this setting, Quantum Neural Networks (QNNs) have gained attention for their potential to revolutionize image classification tasks. Encoding plays a critical role in bridging the gap between classical data and quantum computation, allowing QNNs to utilize quantum resources thereby processing and manipulating classical data in a quantum-mechanical fashion. In this research paper, we present a study of QNNs applied to image classification using the well-known MNIST dataset, with a specific focus on investigating the impact of four quantum encoding methods - Basis, Amplitude, Histogram and Pauli Basis Encoding - on the performance of these QNNs. The research revealed that Basis Encoding and Pauli Basis Encoding outperformed Amplitude Encoding and Histogram Encoding across all the metrics tested. Overally, Pauli Basis Encoding exhibited the best results in enhancing the accuracy and validation loss of the QNNs