Contextual Learning for Stochastic Optimization

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Heuser, Anna, Kesselheim, Thomas
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915298746040320
author Heuser, Anna
Kesselheim, Thomas
author_facet Heuser, Anna
Kesselheim, Thomas
contents Motivated by stochastic optimization, we introduce the problem of learning from samples of contextual value distributions. A contextual value distribution can be understood as a family of real-valued distributions, where each sample consists of a context $x$ and a random variable drawn from the corresponding real-valued distribution $D_x$. By minimizing a convex surrogate loss, we learn an empirical distribution $D'_x$ for each context, ensuring a small Lévy distance to $D_x$. We apply this result to obtain the sample complexity bounds for the learning of an $ε$-optimal policy for stochastic optimization problems defined on an unknown contextual value distribution. The sample complexity is shown to be polynomial for the general case of strongly monotone and stable optimization problems, including Single-item Revenue Maximization, Pandora's Box and Optimal Stopping.
format Preprint
id arxiv_https___arxiv_org_abs_2505_16829
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Contextual Learning for Stochastic Optimization
Heuser, Anna
Kesselheim, Thomas
Machine Learning
Data Structures and Algorithms
Computer Science and Game Theory
Motivated by stochastic optimization, we introduce the problem of learning from samples of contextual value distributions. A contextual value distribution can be understood as a family of real-valued distributions, where each sample consists of a context $x$ and a random variable drawn from the corresponding real-valued distribution $D_x$. By minimizing a convex surrogate loss, we learn an empirical distribution $D'_x$ for each context, ensuring a small Lévy distance to $D_x$. We apply this result to obtain the sample complexity bounds for the learning of an $ε$-optimal policy for stochastic optimization problems defined on an unknown contextual value distribution. The sample complexity is shown to be polynomial for the general case of strongly monotone and stable optimization problems, including Single-item Revenue Maximization, Pandora's Box and Optimal Stopping.
title Contextual Learning for Stochastic Optimization
topic Machine Learning
Data Structures and Algorithms
Computer Science and Game Theory
url https://arxiv.org/abs/2505.16829