Foundations of Federated Learning at the Edge
Abstract
Approved for Public ReleaseThe use of AI-assisted devices is surging in the tactical edge. Training machine learning (ML) models for such applications is, however, a daunting task. There is a substantial amount of data that is collected in a decentralized manner across a large number of edge devices. Federated machine learning (FL) has been proposed as an effective approach to enable decentralized training of machine learning models across edge users. Despite significant recent milestones in FL, there are several fundamental challenges that yet need to be addressed in order to enable its promise. First, edge users have often substantial constraints ontheir resources (e.g., memory, compute, communication, and energy), which severely limits their capability of training large modelslocally. Second, there is large system and data heterogeneity across edge users, which will make their ML objectives and capabilities vastly different. Third, FL clients data distribution continually changewith their interest and new trends. Fourth, as make the ML/AI ecosystem decentralized across edge users, there is potential for more advanced security and privacy breaches into the FL system. The overarching goal of this project is to address these four critical challenges in FL. To that end, the project will consist of the following four integrated thrusts. Thrust 1: Resource-Constrained FL via exploring low-rank structure in models and data.Thrust 2: Adapting to Data and Device Heterogeneity for Deployable Federated Learning.Thrust 3: Federated Continual Learning.Thrust 4: Secure and Resilient Model Aggregation in Federated Learning.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Feb 06, 2023
- Source ID
- N000142312191
Entities
People
- Salman A. Avestimehr
Organizations
- Office of Naval Research
- United States Navy
- University of Southern California