Initially, a project for DS-GA 3001 Tools & Techniques for Machine Learning taught by David Rosenberg. The final repository for that repo is the "original_report" branch.

Published in the 2021 AIMLAI workshop in the ECML PKDD conference.

Paper link: https://arxiv.org/abs/2109.02724

Map of Repository

1. scripts/ice_class.py holds the main ICE feature impact class with replication of ICE plots included
2. notebooks/01-replication/01-ice-plot-from-class generates the Appendix figures in the report and all ICE plots used in the paper
3. notebooks/examples contains examples of ICE feature impact compared to other feature importance/impact metrics on different datasets