Reproduce “Decomposition Based Search - A theoretical and experimental evaluation” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

W.J. van Hoeve, M. Milano

Decomposition Based Search (DBS) was introduced as a generalizable search strategy for constraint satisfaction and optimization problems, demonstrating improved efficiency over Limited Discrepancy Search in finding optimal solutions for Traveling Salesman Problem instances and solutions for Partial Latin Square Completion problems by decoupling subproblem generation and solution.

Key insights

  • DBS can replicate the search behavior of Iterative Broadening (IB) without its inherent redundant re-exploration, offering a more efficient approach for certain problem types.
  • The strategy effectively utilizes 'plateaus' of equally ranked heuristic values, theoretically demonstrating a higher mean probability of success per leaf node compared to LDS in such scenarios.
  • DBS provides a flexible, modular architecture for integrating and exploiting problem-specific heuristics, offering tunable parameters for domain partitioning and subproblem solution strategies.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.