A Preprocessing Perspective for Quantum Machine Learning Classification Advantage in Finance Using NISQ Algorithms.

Area of Science:

• Quantum Computing
• Machine Learning
• Data Science

Background:

• Quantum Machine Learning (QML) has not yet clearly demonstrated advantages over classical approaches.
• Current noisy quantum computers with limited qubits hinder the demonstration of quantum advantage.
• Quantum-inspired techniques show incremental benefits, and hybrid quantum computing shows promise.

Purpose of the Study:

• To enhance classical encoding and performance of quantum classifiers.
• To investigate the impact of Linear Discriminant Analysis (LDA) in data preprocessing for QML.
• To demonstrate improved performance of Variational Quantum Algorithms (VQA) using LDA.

Main Methods:

• Utilizing Linear Discriminant Analysis (LDA) for data preprocessing.
• Implementing and evaluating Variational Quantum Algorithms (VQA) as quantum classifiers.
• Comparing QML performance with and without LDA against baseline classical classifiers.

Main Results:

• LDA preprocessing led to improved classical encoding for quantum classifiers.
• The Variational Quantum Algorithm (VQA) demonstrated a performance gain in balanced accuracy when using LDA.
• The LDA-enhanced VQA model outperformed baseline classical classifiers.

Conclusions:

• Linear Discriminant Analysis (LDA) is an effective technique for improving quantum classifier performance.
• QML methods, when combined with classical preprocessing techniques like LDA, can achieve competitive results.
• Further research into hybrid quantum-classical approaches is warranted to leverage QML potential.