A Collaborative Framework for Quantum Optimisation and Quantum Neural Networks: Credit Feature Selection and Image Classification

TL;DR

This work demonstrates two quantum-assisted ML pipelines: (1) a QUBO-based feature selection approach solved by quantum annealing identifies a compact two-feature subset for credit risk using the German Credit Dataset, achieving 70.0% accuracy with high recall on the low-risk class, and (2) quantum neural networks (CRADL and CRAML) with FRQI-based encodings classify MNIST digits 3 and 6, showing performance on par with classical networks even when qubits are compressed. The study details the QUBO formulation, Ising-machine mapping, and hybrid annealing for feature selection, alongside FRQI and compressed FRQI encodings, wave-function construction, and two QNN architectures tailored for image data. Results indicate that quantum approaches can tackle practical optimization and high-dimensional learning tasks despite current hardware limits, while highlighting the trade-offs between resource expenditure and model efficacy. The findings support the potential of quantum computing for finance and computer vision, and outline concrete avenues for scalability and improved training efficiency in future work.

Abstract

This paper investigates the efficacy of quantum computing in two distinct machine learning tasks: feature selection for credit risk assessment and image classification for handwritten digit recognition. For the first task, we address the feature selection challenge of the German Credit Dataset by formulating it as a Quadratic Unconstrained Binary Optimization (QUBO) problem, which is solved using quantum annealing to identify the optimal feature subset. Experimental results show that the resulting credit scoring model maintains high classification precision despite using a minimal number of features. For the second task, we focus on classifying handwritten digits 3 and 6 in the MNIST dataset using Quantum Neural Networks (QNNs). Through meticulous data preprocessing (downsampling, binarization), quantum encoding (FRQI and compressed FRQI), and the design of QNN architectures (CRADL and CRAML), we demonstrate that QNNs can effectively handle high-dimensional image data. Our findings highlight the potential of quantum computing in solving practical machine learning problems while emphasizing the need to balance resource expenditure and model efficacy.

Figures (9)

• Figure 1: Downsampling and Binarization of Digit '3'
• Figure 2: Quantum circuit for constructing a 4-qubit superposition state.
• Figure 3: Decomposition of an n-qubit controlled gate in terms of (n-1)-qubit controlled gates.
• Figure 4: CRADL Network Structure
• Figure 5: CRAML Network Structure