Controlled importance-weighted cross-validation

This repository contains experiments and visualizations accompanying the paper:

"Robust importance-weighted cross-validation under sample selection bias",

published in the proceedings of the International Workshop on Machine Learning for Signal Processing, 2019 (pdf/preprint).

Contents:

• experiment-1DG contains visualizations throughout the paper.
• experiment-2DG contains experiments on a 2-dimensional Gaussians setting.
• experiment-ozone contains experiments on the UCI Ozone level detection task.
• util contains utility functions and algorithms.

Contact:

Questions and comments can be submitted to the issues tracker.