NAISS
SUPR
NAISS Projects
SUPR
Federated Reinforcement Learning
Dnr:

NAISS 2025/22-1207

Type:

NAISS Small Compute

Principal Investigator:

Kakoli Majumder

Affiliation:

Linköpings universitet

Start Date:

2025-09-08

End Date:

2026-10-01

Primary Classification:

10212: Algorithms

Webpage:

Allocation

Abstract

Federated reinforcement learning enables agents to collaboratively train a global policy without sharing their individual data. However, high communication overhead and probabilistic risk constraints remain critical bottlenecks. These issues become more complex when the number of federated agents increases. We propose algorithms to solve such challenges.