Probabilistic Focal Search Improves AI Bounded-Suboptimal Search Efficiency

VERA Brief

AI-generated. Grounded in the article and its cited sources.

Researchers have introduced Probabilistic Focal Search (PFS), a new algorithm that improves bounded-suboptimal search efficiency in AI. PFS uses a probabilistic approach to guide the search, aiming to reduce the effort and time needed to find feasible solutions by encouraging faster advancement of the lower bound.

Key facts

• Probabilistic Focal Search (PFS) is an advancement over the existing Focal Search (FS) algorithm.
• PFS introduces a probabilistic element to the search process, either following FS guidance or expanding a minimum-f OPEN node.
• This probabilistic approach is intended to encourage the lower bound to advance, potentially leading to feasible solutions more quickly.
• PFS was benchmarked against FS on N-Puzzle, Pancake Sorting, and the Traveling Salesperson Problem (TSP), showing significant reductions in node expansions.
• An anytime extension, Anytime Probabilistic Focal Search (APFS), also demonstrated performance gains over existing anytime algorithms.