Releases: DillonZChen/goose
v2.0.0
Full Changelog: v1.2.2...v2.0.0
ECAI-25
Source code for and extended version of the ECAI-25 publication "Weisfeiler-Leman Features for Planning and Search: A 1,000,000 Sample Size Hyperparameter Study" by Dillon Z. Chen.
v1.2.2
Improved pretrained weights and documentation for pretrained weights
What's Changed
- Improved weights and documentation by @DillonZChen in #21
Full Changelog: v1.2.1...v1.2.2
v1.2.1
Minor quality of life and planning improvements.
- Apptainer that can be installed and pretrained models for common benchmarks (IPC23LT and benchmarks from NeurIPS24 paper)
- Clearer model configuration files
- Change ranking formulation constants to always be 1 (from 0 for maybe pairs), following experiments by @u6684258
- Fix CI tests in GitHub Actions
What's Changed
- Merge dev changes by @DillonZChen in #14
- 16 speed up data extraction for ranking by @DillonZChen in #19
Full Changelog: v1.1.1...v1.2.1
AAAI-24
Source code for the AAAI-24 publication "Learning Domain-Independent Heuristics for Grounded and Lifted Planning" by Dillon Z. Chen, Sylvie Thiébaux, and Felipe Trevizan
v1.1.1
Minor bug fixes.
- Numeric data collection no longer fails silently. The code now exits if an error with NFD occured.
- Update downward to version 24.06
- Add more instructions in the README.
v1.0.1
Clean re-implementation of GOOSE with the WLPlan package. New features include:
- support for classical grounded, classical lifted, and numeric planning
- new ranking formulations
- Apptainer image
- PCA visualisations of training data
Furthermore, support for GNNs is deprecated as they are inferior to WL learned heuristics. We refer to previous releases for GNN implementations.
NeurIPS-24
Source code for the NeurIPS-24 publication "Graph Learning for Numeric Planning" by Dillon Z. Chen and Sylvie Thiébaux
ICAPS-24
Source code for the ICAPS-24 publication "Return to Tradition: Learning Reliable Heuristics with Classical Machine Learning" by Dillon Z. Chen, Felipe Trevizan, and Sylvie Thiébaux