Online Projected Gradient Descent for Stochastic Optimization with Decision-Dependent Distributions

TL;DR

This paper introduces an online stochastic gradient descent approach for tracking solutions in decision-dependent stochastic optimization problems, providing explicit bounds on tracking errors and validating results through electric vehicle charging simulations.

Contribution

It presents a novel online gradient method with theoretical bounds for decision-dependent distributions in stochastic optimization.

Findings

Explicit bounds in expectation and high probability for tracking error.

Validation through numerical simulations in electric vehicle charging.

Effective handling of decision-dependent stochastic costs.

Abstract

This paper investigates the problem of tracking solutions of stochastic optimization problems with time-varying costs that depend on random variables with decision-dependent distributions. In this context, we propose the use of an online stochastic gradient descent method to solve the optimization, and we provide explicit bounds in expectation and in high probability for the distance between the optimizers and the points generated by the algorithm. In particular, we show that when the gradient error due to sampling is modeled as a sub-Weibull random variable, then the tracking error is ultimately bounded in expectation and in high probability. The theoretical findings are validated via numerical simulations in the context of charging optimization of a fleet of electric vehicles.