Constrained Stochastic Optimization

Constrained stochastic optimization tackles the challenge of finding optimal solutions under uncertainty and limitations, aiming to maximize or minimize an objective function while satisfying various constraints. Current research focuses on developing efficient algorithms, such as variations of stochastic gradient methods and evolutionary strategies, to handle complex scenarios including high-dimensional spaces, non-independent data (e.g., Markov chains), and computationally expensive black-box simulators. These advancements are crucial for addressing real-world problems across diverse fields, from fair machine learning and resource allocation to reinforcement learning and engineering design, where optimizing performance under constraints is paramount. The development of robust and efficient methods with convergence guarantees is a key area of ongoing investigation.

Papers

December 7, 2023

Stochastic-Constrained Stochastic Optimization with Markovian Data
Yeongjong Kim, Dabeen Lee
Stochastic Constraint Stochastic Primal Dual Constrained Stochastic Optimization Drift Plus Penalty

November 25, 2023

Multi-fidelity Constrained Optimization for Stochastic Black Box Simulators
Atul Agrawal, Kislaya Ravi, Phaedon-Stelios Koutsourelakis, Hans-Joachim Bungartz
Practical Algorithm Optimization Purpose Constrained Optimization Multi Fidelity Constrained Stochastic Optimization

September 6, 2022

Annealing Optimization for Progressive Learning with Stochastic Approximation
Christos Mavridis, John Baras
Reinforcement Learning Stochastic Approximation Progressive Learning Simulated Annealing Prototype Learning Constrained Stochastic Optimization

May 27, 2022

Statistical Inference of Constrained Stochastic Optimization via Sketched Sequential Quadratic Programming
Sen Na, Michael W. Mahoney
Statistical Inference Nonlinear Programming Quadratic Program Sequential Quadratic Programming Stochastic Constraint Online Statistical Inference Constrained Stochastic Optimization

March 23, 2022

Socially Fair Mitigation of Misinformation on Social Networks via Constraint Stochastic Optimization
Ahmed Abouzeid, Ole-Christoffer Granmo, Christian Webersik, Morten Goodwin
Social Network Misinformation Claim Misinformation Mitigation Fairness Mitigation Constrained Stochastic Optimization

February 21, 2022

A Globally Convergent Evolutionary Strategy for Stochastic Constrained Optimization with Applications to Reinforcement Learning
Youssef Diouane, Aurelien Lucchi, Vihang Patil
Reinforcement Learning Financial Application Optimization Problem Stochastic Optimization Optimization Algorithm Evolution Strategy Constrained Stochastic Optimization