Multitask Online Learning: Listen to the Neighborhood Buzz

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Achddou, Juliette, Cesa-Bianchi, Nicolò, Laforgue, Pierre
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909162719412224
author Achddou, Juliette
Cesa-Bianchi, Nicolò
Laforgue, Pierre
author_facet Achddou, Juliette
Cesa-Bianchi, Nicolò
Laforgue, Pierre
contents We study multitask online learning in a setting where agents can only exchange information with their neighbors on an arbitrary communication network. We introduce $\texttt{MT-CO}_2\texttt{OL}$, a decentralized algorithm for this setting whose regret depends on the interplay between the task similarities and the network structure. Our analysis shows that the regret of $\texttt{MT-CO}_2\texttt{OL}$ is never worse (up to constants) than the bound obtained when agents do not share information. On the other hand, our bounds significantly improve when neighboring agents operate on similar tasks. In addition, we prove that our algorithm can be made differentially private with a negligible impact on the regret. Finally, we provide experimental support for our theory.
format Preprint
id arxiv_https___arxiv_org_abs_2310_17385
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Multitask Online Learning: Listen to the Neighborhood Buzz
Achddou, Juliette
Cesa-Bianchi, Nicolò
Laforgue, Pierre
Machine Learning
We study multitask online learning in a setting where agents can only exchange information with their neighbors on an arbitrary communication network. We introduce $\texttt{MT-CO}_2\texttt{OL}$, a decentralized algorithm for this setting whose regret depends on the interplay between the task similarities and the network structure. Our analysis shows that the regret of $\texttt{MT-CO}_2\texttt{OL}$ is never worse (up to constants) than the bound obtained when agents do not share information. On the other hand, our bounds significantly improve when neighboring agents operate on similar tasks. In addition, we prove that our algorithm can be made differentially private with a negligible impact on the regret. Finally, we provide experimental support for our theory.
title Multitask Online Learning: Listen to the Neighborhood Buzz
topic Machine Learning
url https://arxiv.org/abs/2310.17385