DRO: A Python Library for Distributionally Robust Optimization in Machine Learning

TL;DR

The paper presents dro, an open-source Python library that facilitates distributionally robust optimization in machine learning, supporting multiple formulations and models with significant runtime improvements.

Contribution

Introduces dro, a versatile Python library implementing 14 DRO formulations and 9 models, compatible with scikit-learn and PyTorch, with optimized performance for large datasets.

Findings

Reduces runtime by 10x to over 1000x on large datasets

Supports 79 distinct DRO methods

Compatible with scikit-learn and PyTorch

Abstract

We introduce dro, an open-source Python library for distributionally robust optimization (DRO) for regression and classification problems. The library implements 14 DRO formulations and 9 backbone models, enabling 79 distinct DRO methods. Furthermore, dro is compatible with both scikit-learn and PyTorch. Through vectorization and optimization approximation techniques, dro reduces runtime by 10x to over 1000x compared to baseline implementations on large-scale datasets. Comprehensive documentation is available at https://python-dro.org.