/hxe-for-tda

Primary LanguageJupyter Notebook

DOI

A Weighted Hierarchical Cross-entropy Loss for Classification of Astrophysical Transients

This repository is for the workshop proceeding NeurIPS Workshop paper by Villar et al., and the paper The Impact of Host Galaxy Properties on Supernova Classification with Hierarchical Labels by Villar et al. It is associated with the manuscript number Z.

The supernova-example-24.ipynb notebook contains the most relevant tutorial, and shows classification of a fake set of supernovae using the weighted hierarchical cross-entropy loss.

The supernova_example.ipynb notebook reproduces results for the SN sample from the Zwicky Transient Facility Bright Supernova Survey.

The variable_star_example.ipynb notebook reproduces results for the variable star sample (also from the Zwicky Transient Facility).

Note that python >=3.10 is required.