Simulation-Based Admissible Dominance Pruning
Authors: álvaro Torralba, Jörg Hoffmann
IJCAI 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our experiments show substantial improvements across several IPC benchmark domains. |
| Researcher Affiliation | Academia | Alvaro Torralba and J org Hoffmann Saarland University Saarbr ucken, Germany torralba@cs.uni-saarland.de, hoffmann@cs.uni-saarland.de |
| Pseudocode | Yes | Our algorithm proceeds as follows: For all i, set i:= {(s, t) | s, t Si, s Si G or t Si G} while ex. (i, s, t) s.t. not Ok(i, s, t) do Select one such triple (i, s, t) Set i:= i \{(s, t)} endwhile return R := { 1, . . . , k} |
| Open Source Code | No | The paper states "Our current implementation of that process is largely na ıve." but does not provide any information about open-sourcing the code for the described methodology. |
| Open Datasets | Yes | We ran all optimal-track STRIPS planning instances from the international planning competitions (IPC 98 IPC 14). |
| Dataset Splits | No | The paper uses IPC benchmarks and describes parameters for the merge-and-shrink process (time limit, abstract transition limit M) but does not provide explicit training, validation, or test dataset splits in the traditional machine learning sense (e.g., percentages, sample counts, or specific file paths for splits). |
| Hardware Specification | Yes | All experiments were conducted on a cluster of Intel E5-2660 machines running at 2.20 GHz, with time (memory) cut-offs of 30 minutes (4 GB). |
| Software Dependencies | No | The paper mentions using "Fast Downward (FD) [Helmert, 2006]" but does not provide specific version numbers for FD or any other software dependencies. |
| Experiment Setup | Yes | We ran all optimal-track STRIPS planning instances from the international planning competitions (IPC 98 IPC 14). All experiments were conducted on a cluster of Intel E5-2660 machines running at 2.20 GHz, with time (memory) cut-offs of 30 minutes (4 GB). We run A with FD s blind heuristic, and with LM-cut [Helmert and Domshlak, 2009]. Our initial abstractions are obtained using merge-and-shrink with exact label reduction, bisimulation shrinking, and the non-linear merge DFP strategy [Dr ager et al., 2006; 2009; Sievers et al., 2014]. We impose two bounds on this process, namely a time limit of 300 seconds, as well as a limit M on the number of abstract transitions. A reasonably good magic setting for M, in our current context, is 100k. For M = 0, i. e. computing the component simulations on individual state variables only, performance is substantially worse. For M = 200k, the overhead becomes prohibitive. |