A Unifying View on Individual Bounds and Heuristic Inaccuracies in Bidirectional Search

Authors: Vidal Alcázar, Pat Riddle, Mike Barley2327-2334

AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results show a substantial improvement, up to an order of magnitude in the number of necessarily-expanded nodes compared to state-of-the-art near-optimal algorithms in common benchmarks. Experiments In this section we present an exhaustive experimental evaluation of relevant Bi-HS algorithms.
Researcher Affiliation Academia Vidal Alc azar vidal.alcazar@riken.jp Riken AIP, Tokyo, Japan Pat Riddle, Mike Barley {pat, barley}@cs.auckland.ac.nz University of Auckland, Auckland, New Zealand
Pseudocode Yes Algorithm 1 Main Lower Bound Loop Algorithm 2 Update C
Open Source Code No The paper does not provide an explicit statement about releasing its source code or a link to a code repository for the described methodology.
Open Datasets Yes 14Pancake Puzzle with the GAP heuristic (Helmert 2010). Towers of Hanoi problem with (10+2), (8+4) and (6+6) additive PDBs (Felner, Korf, and Hanan 2004). 15 Sliding Tile Puzzle (Korf 1985) with Manhattan Distance. grid-based pathfinding benchmarks (Sturtevant 2012) with the octile heuristic, using mazes and Dragon Age: Origins (DAO) maps.
Dataset Splits No The paper evaluates search algorithms on problem instances and does not specify explicit training/validation/test dataset splits, which are typical in machine learning contexts.
Hardware Specification No The paper discusses algorithmic performance in terms of "nodes per second" and relative speed differences due to computational complexity but does not provide specific details on the hardware used for experiments (e.g., CPU/GPU models, memory).
Software Dependencies No The paper does not specify any software dependencies with version numbers.
Experiment Setup Yes All algorithms use ϵ if applicable. Initially the f limit is set to max(hf(start), hb(goal)) and both g limits are set to 0. All limits are increased by ι and not by 1.