Online Decentralized Frank-Wolfe: From theoretical bound to applications in smart-building

TL;DR

This paper introduces an online decentralized Frank-Wolfe algorithm for non-convex loss functions, providing theoretical guarantees and demonstrating its effectiveness in smart-building applications.

Contribution

It presents a novel decentralized online optimization algorithm with theoretical bounds, applicable to real-world distributed systems like smart buildings.

Findings

The algorithm achieves convergence guarantees for non-convex losses.

It performs effectively in real-world smart-building scenarios.

Theoretical analysis supports its practical utility.

Abstract

The design of decentralized learning algorithms is important in the fast-growing world in which data are distributed over participants with limited local computation resources and communication. In this direction, we propose an online algorithm minimizing non-convex loss functions aggregated from individual data/models distributed over a network. We provide the theoretical performance guarantee of our algorithm and demonstrate its utility on a real life smart building.