On Parallel External-Memory Bidirectional Search
This work addresses the problem of efficiently solving large-scale search problems for researchers and practitioners in AI and algorithms, representing an incremental advancement by extending bidirectional search to parallel external-memory settings.
The paper tackled the challenge of scaling bidirectional heuristic search to large-scale problems by developing a parallel external-memory variant of the BAE* algorithm (PEM-BAE*). The result showed that PEM-BAE* outperformed existing parallel external-memory variants of A*, the meet-in-the-middle algorithm, and a parallel IDA* algorithm, demonstrating bidirectional search's superiority over unidirectional methods in several domains.
Parallelization and External Memory (PEM) techniques have significantly enhanced the capabilities of search algorithms when solving large-scale problems. Previous research on PEM has primarily centered on unidirectional algorithms, with only one publication on bidirectional PEM that focuses on the meet-in-the-middle (MM) algorithm. Building upon this foundation, this paper presents a framework that integrates both uni- and bi-directional best-first search algorithms into this framework. We then develop a PEM variant of the state-of-the-art bidirectional heuristic search (BiHS) algorithm BAE* (PEM-BAE*). As previous work on BiHS did not focus on scaling problem sizes, this work enables us to evaluate bidirectional algorithms on hard problems. Empirical evaluation shows that PEM-BAE* outperforms the PEM variants of A* and the MM algorithm, as well as a parallel variant of IDA*. These findings mark a significant milestone, revealing that bidirectional search algorithms clearly outperform unidirectional search algorithms across several domains, even when equipped with state-of-the-art heuristics.