Data-driven Optimal Cost Selection for Distributionally Robust Optimization
Authors:
- Jose BlanchetJBJose Blanchet
- Yang KangYKYang Kang
- Fan ZhangFZFan Zhang
- Karthyek MurthyKMKarthyek Murthy
Preprint
Abstract
Recently, (Blanchet, Kang, and Murhy 2016, and Blanchet, and Kang 2017) showed that several machine learning algorithms, such as square-root Lasso, Support Vector Machines, and regularized logistic regression, among many others, can be represented exactly as distributionally robust optimization (DRO) problems. The distributional uncertainty is defined as a neighborhood centered at the empirical distribution. We propose a methodology which learns such neighborhood in a natural data-driven way. We show rigorously that our framework encompasses adaptive regularization as a particular case. Moreover, we demonstrate empirically that our proposed methodology is able to improve upon a wide range of popular machine learning estimators.Keywords
All Related Versions
This publication has 5 references indexed in Scilit:
- Robust Wasserstein profile inference and applications to machine learningJournal of Applied Probability, 2019
- Data-driven distributionally robust optimization using the Wasserstein metric: performance guarantees and tractable reformulationsMathematical Programming, 2017
- A Mahalanobis metric learning-based polynomial kernel for classification of hyperspectral imagesNeural Computing and Applications, 2016
- Geometry and properties of generalized ridge regression in high dimensions2014
- The Adaptive Lasso and Its Oracle PropertiesJournal of the American Statistical Association, 2006