Stochastic Optimization under Hidden Convexity


METADATA ONLY
Loading...

Date

2023-12-15

Publication Type

Conference Paper

ETH Bibliography

yes

Citations

Altmetric
METADATA ONLY

Data

Rights / License

Abstract

In this work, we consider constrained stochastic optimization problems under hidden convexity, i.e., those that admit a convex reformulation via non-linear (but invertible) map c : X → U. A number of non-convex problems ranging from optimal control, revenue and inventory management, to convex reinforcement learning all admit such a hidden convex structure. Unfortunately, in the majority of applications considered, the map c(⋅) is unavailable or implicit; therefore, directly solving the convex reformulation is not possible. On the other hand, the stochastic gradients with respect to the original variable x ϵ X are often easy to obtain. Motivated by these observations, we examine the basic projected stochastic (sub-) gradient methods for solving such problems under hidden convexity. We provide the first sample complexity guarantees for global convergence in smooth and non-smooth settings. Additionally, in the smooth setting, we improve our results to the last iterate convergence in terms of function value gap using the momentum variant of projected stochastic gradient descent.

Publication status

published

Editor

Book title

OPT 2023: Optimization for Machine Learning (NeurIPS 2023 Workshop)

Journal / series

Volume

Pages / Article No.

Publisher

OpenReview

Event

15th International OPT Workshop on Optimization for Machine Learning (OPT 2023) held at NeurIPS 2023

Edition / version

Methods

Software

Geographic location

Date collected

Date created

Subject

Optimization and Control (math.OC); Computational Complexity (cs.CC); FOS: Mathematics; FOS: Computer and information sciences; 90C06, 90C15, 90C26; hidden convexity; stochastic optimization; global convergence

Organisational unit

09729 - He, Niao / He, Niao check_circle

Notes

Poster presentation on December 15, 2023.

Funding

Related publications and datasets

Is new version of: 10.48550/arXiv.2401.00108