A preprocessing perspective for quantum machine learning classification advantage using NISQ algorithms

TL;DR

This paper demonstrates that applying Linear Discriminant Analysis during data preprocessing can enhance the performance of quantum classifiers, potentially providing advantages over classical methods in noisy intermediate-scale quantum (NISQ) devices.

Contribution

It introduces a preprocessing technique using LDA to improve quantum classifier performance, highlighting a practical approach to achieving quantum advantage.

Findings

LDA preprocessing improves quantum classifier accuracy.

Quantum classifiers outperform classical baselines with LDA.

Performance gains are notable on NISQ devices.

Abstract

Quantum Machine Learning (QML) hasn't yet demonstrated extensively and clearly its advantages compared to the classical machine learning approach. So far, there are only specific cases where some quantum-inspired techniques have achieved small incremental advantages, and a few experimental cases in hybrid quantum computing are promising considering a mid-term future (not taking into account the achievements purely associated with optimization using quantum-classical algorithms). The current quantum computers are noisy and have few qubits to test, making it difficult to demonstrate the current and potential quantum advantage of QML methods. This study shows that we can achieve better classical encoding and performance of quantum classifiers by using Linear Discriminant Analysis (LDA) during the data preprocessing step. As a result, Variational Quantum Algorithm (VQA) shows a gain of performance in balanced accuracy with the LDA technique and outperforms baseline classical classifiers.