Maximisation of Admissible Multi-Objective Heuristics

Authors: Patrik Haslum, Ryan Xiao Wang

JAIR 2023 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The conclusion of our experimental results is that somax achieves a trade-off similar to that intended with admax cheaper to compute but less informed also when compared to an improved comax implementation.
Researcher Affiliation Academia Patrik Haslum EMAIL Ryan Xiao Wang EMAIL The Australian National University, Canberra, Australia
Pseudocode Yes Appendix A.3 Algorithm (1) Assume V1 = v1, v2, . . . , v|V1| and V2 = u1, u2, . . . , u|V2| are individually sorted in lexicographic order. Initialise p = q = 1 (p and q are indices into V1 and V2, respectively). (2) While p |V1| and q |V2|: (2.1) If vp = uq, add it to Result (by condition (i)) and increment both p and q by 1.
Open Source Code No The paper mentions using "their MO planner implementation, modifying only the MO maximum operators" from Geißer et al. (2022), but does not provide specific access to the authors' own modified code or implementation.
Open Datasets Yes We compare the use of comax and somax in the canonical MO PDB, the MO Hmax and MO H2 heuristics defined by Geißer et al. (2022) on their collection of 437 benchmark MO planning problems.
Dataset Splits No The paper mentions evaluating on a "collection of 437 benchmark MO planning problems" but does not provide any specific information regarding training, validation, or test dataset splits.
Hardware Specification No The paper does not provide any specific hardware details such as GPU models, CPU types, or memory specifications used for running its experiments.
Software Dependencies No The paper does not provide specific version numbers for any software components, libraries, or operating systems used in the experiments.
Experiment Setup Yes We use their MO planner implementation, modifying only the MO maximum operators, and like them we use all non-redundant patterns of 2 or 3 variables for the PDB heuristic.