Beliefs We Can Believe in: Replacing Assumptions with Data in Real-Time Search
Authors: Maximilian Fickert, Tianyi Gu, Leonhard Staut, Wheeler Ruml, Joerg Hoffmann, Marek Petrik9827-9834
AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results show that DDNancy can perform well in domains in which the original assumption-based Nancy performs poorly. |
| Researcher Affiliation | Academia | 1Saarland University, Saarland Informatics Campus, Saarbr ucken, Germany 2Department of Computer Science, University of New Hampshire, USA |
| Pseudocode | Yes | Algorithm 1: Nancy; Algorithm 2: Risk-Based Lookahead; Algorithm 3: Risk Value of a Top-Level Action TLA |
| Open Source Code | No | The paper does not provide an explicit statement about releasing source code for the described methodology or a link to a code repository. |
| Open Datasets | Yes | We used h LM-cut (Helmert and Domshlak 2009) due to it being a popular admissible heuristic... ...We ran the experiments on a cluster of Intel Xeon E5-2660 machines using the lab framework (Seipp et al. 2017). ...We also evaluate our algorithms on the classic 100 15-puzzle instances published by Korf (1985). |
| Dataset Splits | No | The paper describes an 'offline training phase' for generating data for DDNancy's belief distributions and mentions evaluating on instances, but does not specify explicit training/validation/test splits for the models evaluated in the main experiments. |
| Hardware Specification | Yes | We ran the experiments on a cluster of Intel Xeon E5-2660 machines |
| Software Dependencies | No | The paper mentions software like 'Fast Downward' and 'lab framework' along with their respective foundational papers, but does not provide specific version numbers for any software dependencies. |
| Experiment Setup | Yes | We test lookaheads of size 100, 300, and 1,000 nodes. ... used the standard limits of 30 minutes and 4 GB memory in all experiments. ... we used weighted A* with a weight of 2 and the same heuristic function that will be used online. |