PRObabilistic CLAssification Metrics
This is a space for developing the metrics for the Photometric LSST AStronomical TIme Series Classification Challenge (PLAsTiCC), and, generically, other use cases of classification probabilities.
This repository's immediate goals were to implement the metrics described here and demonstrate them on mock classification results, in an effort to identify the vulnerabilities and strengths of the metrics under consideration for the Kaggle competition. This repository served not only as a space for experimenting with how to combine metrics for the competition's win condition but also for testing further metrics for science-specific papers.
The code thus has a flexible, modular architecture that can be recycled for future competitions with different science goals.
To get involved, check out the README
s in the code directories.
Note: Contributors to this repository were not disqualified from competing in the PLAsTiCC.
To use this code in an academic publication, cite doi:10.5281/zenodo.3352638 and doi:10.3847/1538-3881/ab3a2f.
- Alex Malz (GCCL@RUB)
- Tarek Allam (UCL)
- Anita Bahmanyar (U Toronto)
- Rahul Biswas (U Stockholm)
- Renee Hlozek (U Toronto)
- Emille Ishida (CNRS LPC)
- Rafael Martinez-Galarza (CfA)
- Gautham Narayan (UIUC)
- Kara Ponder (UC Berkeley)
- Christian Setzer (U Stockholm)
- The PLAsTiCC Team