On the robust learning mixtures of linear regressions

Ying Huang,Liang Chen

DOI: https://doi.org/10.48550/arXiv.2305.15317

2023-05-23

Abstract:In this note, we consider the problem of robust learning mixtures of linear regressions. We connect mixtures of linear regressions and mixtures of Gaussians with a simple thresholding, so that a quasi-polynomial time algorithm can be obtained under some mild separation condition. This algorithm has significantly better robustness than the previous result.

Machine Learning