A Result on the Computational Complexity of Heuristic Estimates for the A* Algorithm
Abstract
The performance of a new heuristic search algorithm is analyzed. The algorithm uses a formal representation (semantic representation) that contains enough information to compute the heuristic evaluation function h(n), as defined in the context of A∗ , without requiring a human expert to provide it. The heuristic is computed by solving less constrained subproblems (auxiliary problems) of the given problem. The new algorithm is shown to be less efficient than the Dijkstra algorithm, according to the complexity measure “number of node expansions.” This proves that it is not efficient to compute heuristics for A∗ by solving auxiliary problems with backtracking.
Cite
Text
Valtorta. "A Result on the Computational Complexity of Heuristic Estimates for the A* Algorithm." International Joint Conference on Artificial Intelligence, 1983. doi:10.1016/0020-0255(84)90009-4Markdown
[Valtorta. "A Result on the Computational Complexity of Heuristic Estimates for the A* Algorithm." International Joint Conference on Artificial Intelligence, 1983.](https://mlanthology.org/ijcai/1983/valtorta1983ijcai-result/) doi:10.1016/0020-0255(84)90009-4BibTeX
@inproceedings{valtorta1983ijcai-result,
title = {{A Result on the Computational Complexity of Heuristic Estimates for the A* Algorithm}},
author = {Valtorta, Marco},
booktitle = {International Joint Conference on Artificial Intelligence},
year = {1983},
pages = {777-779},
doi = {10.1016/0020-0255(84)90009-4},
url = {https://mlanthology.org/ijcai/1983/valtorta1983ijcai-result/}
}