Debiasing data: An efficient framework using the principle of maximum entropy

In a new paper published at ICML 2020, researchers affiliated with the CSI show how the max-entropy framework, a well-studied idea from statistical physics, leads to a new, efficient algorithm for mitigating social biases in data. To read more, see the blog post and full paper below.

ICML paper: https://proceedings.icml.cc/static/paper_files/icml/2020/2750-Paper.pdf