Optimal Auctions through Deep Learning: Advances in Differentiable Economics

TL;DR

This paper explores the use of deep learning to automate the design of optimal auctions, successfully recovering known solutions and discovering new mechanisms in complex multi-item settings.

Contribution

It introduces a neural network framework for auction design, framing it as a constrained learning problem, and demonstrates its effectiveness through extensive experiments.

Findings

Recovered all known optimal auction solutions

Discovered novel auction mechanisms for complex settings

Provided generalization bounds for the neural network approach

Abstract

Designing an incentive compatible auction that maximizes expected revenue is an intricate task. The single-item case was resolved in a seminal piece of work by Myerson in 1981, but more than 40 years later a full analytical understanding of the optimal design still remains elusive for settings with two or more items. In this work, we initiate the exploration of the use of tools from deep learning for the automated design of optimal auctions. We model an auction as a multi-layer neural network, frame optimal auction design as a constrained learning problem, and show how it can be solved using standard machine learning pipelines. In addition to providing generalization bounds, we present extensive experimental results, recovering essentially all known solutions that come from the theoretical analysis of optimal auction design problems and obtaining novel mechanisms for settings in which the optimal mechanism is unknown.