Layer-wise QUBO-Based Training of CNN Classifiers for Quantum Annealing

Researchers propose a new quantum annealing framework for training CNN classifiers that avoids gradient-based optimization by using Quadratic Unconstrained Binary Optimization (QUBO). The method shows competitive performance with classical approaches on image classification benchmarks while remaining compatible with current D-Wave quantum hardware.

• →New QUBO-based framework enables CNN training via quantum annealing without gradient-based circuit optimization.
• →Method splits classification problems into independent QUBOs with problem size dependent on image resolution and bit precision, not training samples.
• →20-bit formulation matches or exceeds classical gradient descent performance on MNIST, Fashion-MNIST, and EMNIST datasets.
• →Framework remains within qubit and coupler limits of current D-Wave Advantage quantum hardware at 15-bit precision.
• →Approach addresses scalability issues in quantum kernel methods and barren plateau problems in variational quantum circuits.