AI on Quantum Annealers: Practical Quantum Machine Learning

Implementing artificial intelligence and machine learning (AI/ML) on quantum computers has the potential to be transformative — advancing science and technology, enabling the design of new materials and drugs, and improving risk analysis, exploration, and mission planning. However, realizing AI/ML on gate-based quantum computers remains challenging due to their limited number of qubits (typically only a few hundred) and the high error rates inherent in current quantum hardware. In our research, we focus on developing AI/ML algorithms for quantum annealing computers, which are currently more mature and provide access to approximately 5,000 qubits. Using quantum optimization for a feed-forward neural network, we designed and trained a visual recognition model for binary image classification. The quantum-trained model successfully classified images from the MNIST dataset as well as several healthcare-related datasets. We further present benchmarking results of the model across multiple quantum annealer hardware architectures.