Abstract:Superquantiles have recently gained significant interest as a risk-aware metric for addressing fairness and distribution shifts in statistical learning and decision making problems. This paper introduces a fast, scalable and robust second-order computational framework to solve large-scale optimization problems with superquantile-based constraints. Unlike empirical risk minimization, superquantile-based optimization requires ranking random functions evaluated across all scenarios to compute the tail conditional expectation. While this tail-based feature might seem computationally unfriendly, it provides an advantageous setting for a semismooth-Newton-based augmented Lagrangian method. The superquantile operator effectively reduces the dimensions of the Newton systems since the tail expectation involves considerably fewer scenarios. Notably, the extra cost of obtaining relevant second-order information and performing matrix inversions is often comparable to, and sometimes even less than, the effort required for gradient computation. Our developed solver is particularly effective when the number of scenarios substantially exceeds the number of decision variables. In synthetic problems with linear and convex diagonal quadratic objectives, numerical experiments demonstrate that our method outperforms existing approaches by a large margin: It achieves speeds more than 750 times faster for linear and quadratic objectives than the alternating direction method of multipliers as implemented by OSQP for computing low-accuracy solutions. Additionally, it is up to 25 times faster for linear objectives and 70 times faster for quadratic objectives than the commercial solver Gurobi, and 20 times faster for linear objectives and 30 times faster for quadratic objectives than the Portfolio Safeguard optimization suite for high-accuracy solution computations.

Computing moment inequality models using constrained optimization

The Linear Programming of Uncertain System with Interval Inequation Constraint Coefficients

Introducing Moment: A toolkit for semi-definite programming with moment matrices

Inference on estimators defined by mathematical programming

A New Methodology for Robust Optimizations of Optimal Design Problems under Interval Uncertainty

Constraint Handling in Bayesian Optimization -- A Comparative Study of Support Vector Machine, Augmented Lagrangian and Expected Feasible Improvement

Distributionally Robust Optimization Under Moment Uncertainty with Application to Data-Driven Problems

Optimality-based domain reduction for inequality-constrained NLP and MINLP problems

Multinomial Models with Linear Inequality Constraints: Overview and Improvements of Computational Methods for Bayesian Inference

Portfolio selection under model uncertainty: a penalized moment-based optimization approach

A Moment-SOS Hierarchy for Robust Polynomial Matrix Inequality Optimization with SOS-Convexity

A short note on solving box inequality and linear equality constrained optimization problem

Optimal H2 moment matching-based model reduction for linear systems by (non)convex optimization

Constructive approaches to concentration inequalities with independent random variables

Unconstrained Optimization Method for Solving Linear Inequalities

A generalized moment approach to sharp bounds for conditional expectations

Distributionally Robust Scheduling on Parallel Machines under Moment Uncertainty

Practical globally optimal consensus maximization by Branch-and-bound based on interval arithmetic

A cutting surface algorithm for semi-infinite convex programming with an application to moment robust optimization

Fast Computation of Superquantile-Constrained Optimization Through Implicit Scenario Reduction

A fast computational method for moment-independent uncertainty importance measure.