namkoong-lab/whyshift

A python package providing a benchmark with various specified distribution shift patterns.

41
/ 100
Emerging

Researchers and data scientists building machine learning models need to understand how well their models perform when the underlying data patterns shift. This tool provides a benchmark with real-world tabular datasets like income, public health insurance, mobility, taxi trip durations, and accident severity. It helps evaluate how robust a model is to various specified distribution shifts, giving insights into its reliability beyond the training environment. The primary users are researchers focused on fair machine learning, robust AI, and out-of-distribution generalization.

No commits in the last 6 months. Available on PyPI.

Use this if you are a researcher who needs to rigorously test your machine learning models against known, controlled distribution shifts on diverse real-world datasets to understand their generalization capabilities.

Not ideal if you are looking for a general-purpose data preprocessing library or a tool to automatically correct for arbitrary data shifts without defined patterns.

Machine Learning Research Fair AI Robustness Testing Out-of-Distribution Generalization Predictive Modeling
Stale 6m
Maintenance 0 / 25
Adoption 8 / 25
Maturity 25 / 25
Community 8 / 25

How are scores calculated?

Stars

59

Forks

4

Language

Jupyter Notebook

License

Last pushed

Nov 27, 2023

Commits (30d)

0

Dependencies

9

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/namkoong-lab/whyshift"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.