Block-Level Goal Recognition Design
Authors: Tsz-Chiu Au
AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conducted two experiments to evaluate the pruning rules and compare the BFS with a local search algorithm... In Table 1, we can see that our BFS with prunedreduce and design subtree pruning outperforms the original BFS with pruned-reduce by around an order of magnitude. |
| Researcher Affiliation | Academia | Tsz-Chiu Au Department of Computer Science and Engineering, Ulsan National Institute of Science and Technology chiu@unist.ac.kr |
| Pseudocode | Yes | The pseudocode of the BFS can be found in the technical appendix... The pseudocode of the BFS with the design subtree pruning rule can be found in the technical appendix... The local search algorithm s pseudocode can be found in the technical appendix. |
| Open Source Code | No | The paper does not include any explicit statement about releasing source code or a link to a code repository. |
| Open Datasets | No | We adopted four domains in the International Planning Competition: LOGISTICS, GRID, DEPOTS, and DRIVERLOG... we implemented a problem generator that can also generate a hierarchical design model for each problem instance." The paper describes generating problem instances but does not provide concrete access (link, DOI, repository, or specific citation for the *generated datasets*) to the data used in experiments. |
| Dataset Splits | No | The paper does not provide specific dataset split information (exact percentages, sample counts, citations to predefined splits, or detailed splitting methodology) needed to reproduce the data partitioning. The problem is a search/design problem, not a typical machine learning training task. |
| Hardware Specification | Yes | Both experiments were conducted on an Apple laptop with an M1 CPU and 16GB RAM. |
| Software Dependencies | No | The paper mentions using 'Fast Downward (Helmert 2006)' but does not provide specific version numbers for this or any other software dependencies. |
| Experiment Setup | Yes | In each domain, we generated 30 problem instances of different sizes, and for each problem instance, we generated n goals, where 2 n 10 and n increases with the problem size. Then, we used a planner called Fast Downward (Helmert 2006) to find a legal plan and the corresponding legal path for each goal. We randomly deleted some actions on the legal plans and ran Fast Downward again to find legal plans that were slightly different from the previous ones. Ultimately, the number of legal paths in P leg was around 5 n. |