aimalz/proclam

Real classifier systematics

aimalz opened this issue · 1 comments

It would be valuable to test our metrics on confusion matrices and/or real classification probabilities or deterministically predicted classes (plus true class assignments), of any set of transient types. @reneehlozek started this in #27, and @AshishMahabal has committed to continuing it.

GitHub has a limit on storage space, so please put any files in a public place on the web where the pipeline notebook can access them, especially if they're larger than 1 MB. If for some reason you can't do this, contact me and we'll work something out.

The file format here would be best, but it's not a dealbreaker. If the files are not straightforward plaintext, please include a demonstration of how to read in the data in pipeline_sandbox.ipynb or provide a README with instructions.