Nonnegative/Binary matrix factorization for image classification using quantum annealing.

Area of Science:

• Quantum Computing
• Machine Learning
• Image Classification

Background:

• Machine learning (ML) has advanced significantly with classical computing.
• Integrating quantum technology, specifically quantum computing, promises substantial benefits and advancements in computational capabilities.
• Quantum annealing is an emerging quantum computing technique being explored for various ML applications.

Purpose of the Study:

• To implement a matrix factorization method using quantum annealing for image classification.
• To compare the performance of this quantum-enhanced method against traditional ML techniques.
• To demonstrate the practical benefits of integrating quantum annealing with machine learning.

Main Methods:

• Implementation of a Nonnegative/binary matrix factorization (NBMF) model, originally a generative model, for multiclass classification.
• Feature extraction from handwritten digit images using NBMF.
• Application of extracted features to solve image classification problems.
• Comparison with classical ML methods like neural networks.

Main Results:

• Quantum annealing-based NBMF demonstrated superior accuracy for image classification compared to classical methods when data, features, and epochs were limited.
• Training ML models using a quantum annealing solver significantly reduced computation time.
• The study confirmed benefits of using quantum annealing technology with machine learning under specific conditions.

Conclusions:

• Quantum annealing offers a promising approach to enhance machine learning, particularly for image classification tasks.
• The integration of quantum annealing can lead to more efficient and accurate ML models, especially in resource-constrained scenarios.
• This research highlights the potential of quantum computing to revolutionize current machine learning paradigms.