Efficient Algorithms for Empirical Group Distributionally Robust Optimization and Beyond
This work addresses optimization efficiency for robust machine learning models across multiple groups, representing an incremental improvement in algorithmic design.
The paper tackles the empirical Group Distributionally Robust Optimization (GDRO) problem by developing the ALEG algorithm, which achieves an ε-accuracy computation complexity of O(m√(n̄ ln m)/ε), outperforming state-of-the-art methods by a factor of √m, and extends this to the Minimax Excess Risk Optimization (MERO) problem with similar efficiency gains.
In this paper, we investigate the empirical counterpart of Group Distributionally Robust Optimization (GDRO), which aims to minimize the maximal empirical risk across $m$ distinct groups. We formulate empirical GDRO as a $\textit{two-level}$ finite-sum convex-concave minimax optimization problem and develop an algorithm called ALEG to benefit from its special structure. ALEG is a double-looped stochastic primal-dual algorithm that incorporates variance reduction techniques into a modified mirror prox routine. To exploit the two-level finite-sum structure, we propose a simple group sampling strategy to construct the stochastic gradient with a smaller Lipschitz constant and then perform variance reduction for all groups. Theoretical analysis shows that ALEG achieves $\varepsilon$-accuracy within a computation complexity of $\mathcal{O}\left(\frac{m\sqrt{\bar{n}\ln{m}}}{\varepsilon}\right)$, where $\bar n$ is the average number of samples among $m$ groups. Notably, our approach outperforms the state-of-the-art method by a factor of $\sqrt{m}$. Based on ALEG, we further develop a two-stage optimization algorithm called ALEM to deal with the empirical Minimax Excess Risk Optimization (MERO) problem. The computation complexity of ALEM nearly matches that of ALEG, surpassing the rates of existing methods.