Reverse Annealing for Nonnegative/Binary Matrix Factorization
This work addresses a specific bottleneck in quantum computing applications for matrix factorization, offering an incremental improvement in algorithm performance.
The paper tackled the performance plateau in quantum annealing for nonnegative/binary matrix factorization by using reverse annealing to refine solutions after an initial global search, resulting in significant improvements for most run times compared to forward annealing alone.
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.