Reverse Annealing for Nonnegative/Binary Matrix Factorization

TL;DR

This paper explores the use of reverse annealing in quantum annealing algorithms to improve nonnegative and binary matrix factorization, achieving better performance than traditional forward annealing.

Contribution

It introduces reverse annealing as a refinement step after initial global search, enhancing quantum annealing's effectiveness in matrix factorization tasks.

Findings

Reverse annealing improves solution quality over forward annealing.

Combining forward and reverse annealing outperforms forward annealing alone.

Performance gains are significant except for very short run times.

Abstract

It was recently shown that quantum annealing can be used as an effective, fast subroutine in certain types of matrix factorization algorithms. The quantum annealing algorithm performed best for quick, approximate answers, but performance rapidly plateaued. In this paper, we utilize reverse annealing instead of forward annealing in the quantum annealing subroutine for nonnegative/binary matrix factorization problems. After an initial global search with forward annealing, reverse annealing performs a series of local searches that refine existing solutions. The combination of forward and reverse annealing significantly improves performance compared to forward annealing alone for all but the shortest run times.