The paper introduces a multilevel algorithm reinforced with a spectral graph representation learning-based accelerator and quantum-informed recursive optimization to tackle large-scale graph maximum cut instances. The key highlights are:
The multilevel approach decomposes the original problem into a hierarchy of progressively simpler, related sub-problems at coarser levels, which are more feasible for the currently available quantum hardware.
The graph representation learning model utilizes the idea of QAOA variational parameters concentration to substantially improve the performance of QAOA on the sub-problems.
The quantum-informed recursive optimization algorithm leverages quantum information to derive potential classical problem-specific reductions, recursively simplifying the original problem.
The experimental results demonstrate the potential of using the proposed multilevel approaches on very large graphs by achieving high-quality solutions in a much faster time compared to previous hybrid quantum-classical decomposition-based algorithms.
The reinforced multilevel scheme outperforms classical state-of-the-art solvers on diverse sets of graphs, including real-world problems, optimization instances, and graphs that are hard for the Goemans-Williamson MAXCUT approximation.
To Another Language
from source content
arxiv.org
Thông tin chi tiết chính được chắt lọc từ
by Bao Bach,Jos... lúc arxiv.org 04-23-2024
https://arxiv.org/pdf/2404.14399.pdfYêu cầu sâu hơn