TY - GEN
T1 - Improving bidirectional heuristic search by bounds propagation
AU - Shperberg, Shahaf S.
AU - Felner, Ariel
AU - Sturtevant, Nathan R.
AU - Shimony, Solomon E.
AU - Hayoun, Avi
N1 - Publisher Copyright:
Copyright © 2019, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved.
PY - 2019/1/1
Y1 - 2019/1/1
N2 - Recent research on bidirectional search describes anomalies, or cases in which improved heuristics lead to more node expansions. Aiming to avoid such anomalies, this paper characterizes desirable properties for bidirectional search algorithms, and studies conditions for obtaining these properties. The characterization is based on a recently developed theory for bidirectional search, which has formulated conditions on pairs of nodes such that at least one node from every pair meeting these conditions must be expanded. Moreover, based on this must-expand-pairs theory, we introduce a method for enhancing heuristics by propagating lower bounds (lb-propagation) between frontiers. This lb-propagation can bestow the desirable properties on some existing algorithms (e.g., the MM family) while avoiding the above anomaly altogether. Empirical results show that lb-propagation reduces the number of node expansions in many cases.
AB - Recent research on bidirectional search describes anomalies, or cases in which improved heuristics lead to more node expansions. Aiming to avoid such anomalies, this paper characterizes desirable properties for bidirectional search algorithms, and studies conditions for obtaining these properties. The characterization is based on a recently developed theory for bidirectional search, which has formulated conditions on pairs of nodes such that at least one node from every pair meeting these conditions must be expanded. Moreover, based on this must-expand-pairs theory, we introduce a method for enhancing heuristics by propagating lower bounds (lb-propagation) between frontiers. This lb-propagation can bestow the desirable properties on some existing algorithms (e.g., the MM family) while avoiding the above anomaly altogether. Empirical results show that lb-propagation reduces the number of node expansions in many cases.
UR - http://www.scopus.com/inward/record.url?scp=85074952935&partnerID=8YFLogxK
M3 - Conference contribution
AN - SCOPUS:85074952935
T3 - Proceedings of the 12th International Symposium on Combinatorial Search, SoCS 2019
SP - 106
EP - 114
BT - Proceedings of the 12th International Symposium on Combinatorial Search, SoCS 2019
A2 - Surynek, Pavel
A2 - Yeoh, William
PB - AAAI press
T2 - 12th International Symposium on Combinatorial Search, SoCS 2019
Y2 - 16 July 2019 through 17 July 2019
ER -