Prior-Free Double Auctions in Incentive Design for Federated Learning

Date: 

Friday, March 26, 2021, 10:00am to 11:30am

Location: 

Zoom conference

PresenterAndreas Haupt, PhD Student with the MIT Institute for Data Systems and Society
Topic: Prior-Free Double Auctions in Incentive Design for Federated Learning

Spring 2021 Seminar

The EconCS Group holds an Economics and Computer Science research seminar each semester.

Spring 2021 meetings are held at 10-11:30am on Fridays. Seminar Coordinators for Spring '21 are Mark York, markyork@g.harvard.edu and Anson Kang, ansonkahng@college.harvard.edu
 

Abstract: Federated learning (FL) is a paradigm that allows distributed clients to learn a shared machine learning model without sharing their sensitive training data. While FL is successfully applied in environments where a central orchestrator has the primary interest in the shared model’s quality, incentive problems complicate FL in settings with self-interested, heterogenous clients. In this talk, we model FL incentive design as a digital goods double auction with a submodular production function. Inspired by existing prior-free auction designs, we present a prior-free mechanism that constant-factor approximates optimal revenue . We introduce how “model leakage”, a measure of model sharing between clients beyond the control of the designer, impacts mechanism performance. To facilitate further study, we present code that allows to test incentive designs for federated learning using clients training on data splits from the MNIST, Fashion-MNIST and CIFAR-10 datasets. A potential broader impact of this research might be that, for standard learning tasks, incentivized FL might allow for privacy-preserving, sustainable and scalable data transactions. Joint work with V. Mugunthan

Speaker Bio: Andreas Haupt is a graduate student at the Institute for Data, Systems, and Society (IDSS), where he works with Munther Dahleh and Alessandro Bonatti. Prior to joining IDSS, Andreas received master’s degrees in Mathematics and Economics and bachelor’s degrees in Computer Science and Mathematics from the universities of Bonn and Frankfurt. In 2019, he was an MIT Presidential Fellow. Andreas conducts research on the role of communication and complexity in market design and is interested in designs of markets for information. He has interned in digital policy for the German federal parliament and the European Union’s competition authorities for digital markets.