Automatic Configuration of Sequential Planning Portfolios
Authors: Jendrik Seipp, Silvan Sievers, Malte Helmert, Frank Hutter
AAAI 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate Cedalion empirically by applying it to construct sequential planning portfolios based on component planners from the highly parameterized Fast Downward (FD) framework. Results for a broad range of planning settings demonstrate that without any knowledge of planning or FD Cedalion constructs sequential FD portfolios that rival, and in some cases substantially outperform, manually-built FD portfolios. |
| Researcher Affiliation | Academia | Jendrik Seipp, Silvan Sievers, Malte Helmert University of Basel Basel, Switzerland {jendrik.seipp,silvan.sievers,malte.helmert}@unibas.ch Frank Hutter University of Freiburg Freiburg, Germany fh@cs.uni-freiburg.de |
| Pseudocode | Yes | Algorithm 1 : Cedalion. |
| Open Source Code | No | The paper does not provide an explicit statement or link for the open-source code of the Cedalion algorithm. It mentions listing 'found portfolios' in a separate technical report, which refers to experimental results, not the source code. |
| Open Datasets | No | The paper describes using 'instances of the corresponding IPC 2011 track' and 'tasks that Fawcett et al. (2011) generated' for training. While these are recognized benchmarks in the planning community, the paper does not provide specific links, DOIs, repositories, or formal citations for *accessing* these exact datasets, which is required by the prompt. |
| Dataset Splits | No | The paper describes training and test sets and their construction (stratified sampling) but does not mention the use of a separate validation set or provide details for such a split. |
| Hardware Specification | No | The paper mentions a general 'CPU limits' and a memory limit ('2 GB of memory') but does not specify any particular CPU or GPU models, memory amounts, or details of the computational hardware used for the experiments. |
| Software Dependencies | No | The paper names software used (e.g., Fast Downward, SMAC, Param ILS, Clasp) but does not provide specific version numbers for any of these software components. |
| Experiment Setup | Yes | We used 10 parallel SMAC runs in each iteration of Cedalion and chose the configuration, runtime pair that maximized the performance metric on the current instance set. Each of the SMAC runs was given a wall-clock time budget of 10h in all settings except optimal planning, where we used 30h to account for the fact that optimal planning is generally harder than satisficing planning. |