Capabilities of an adiabatic quantum computer for pattern recognition

This study compares three discrete-variable models: the fully connected Ising model, the quadratic unconstrained binary optimization model, and the Hopfield neural network. The analysis evaluates their potential interchangeability in solving optimization and pattern recognition tasks, as well as their applicability to quantum computing. A novel method for pattern storage and recognition is proposed, based on the dynamics of forward and reverse quantum annealing. This method has been successfully tested on a D-Wave adiabatic quantum computer. An optimal annealing parameter was empirically determined, maximizing recognition accuracy while minimizing annealing time. The study also highlights limitations related to the size and noise levels of the patterns, imposed by the quantum processor's architectural constraints. Finally, conclusions are drawn regarding the effectiveness implications of the proposed approach.