A Distributionally Robust Optimization Approach for Outlier Detection

Ruidi Chen1, Ioannis Ch. Paschalidis1

  • 1Boston University



Regular Session


10:00 - 12:00 | Mon 17 Dec | Glimmer 2 | MoA10

Machine Learning I

Full Text


We consider the outlier detection problem in a linear regression setting. Outlying observations can be detected by large residuals but this approach is not robust to large outliers which tend to shift the residual function. Instead, we propose a new Distributionally Robust Optimization (DRO) method addressing this issue. The robust optimization problem reduces to solving a second-order cone programming problem. We prove several generalization guarantees for our solution under mild conditions. Extensive numerical experiments demonstrate that our approach outperforms Huber's robust regression approach.

Additional Information

No information added


No videos found