Learning Team Decisions
Abstract
In this paper, we treat linear quadratic team decision problems, where a team of agents minimizes a convex quadratic cost function over time steps subject to possibly distinct linear measurements of the state of nature. We assume that the state of nature is a Gaussian random variable and that the agents do not know the cost function nor the linear functions mapping the state of nature to their measurements. We present a gradient-descent based algorithm with an expected regret of for full information gradient feedback and for bandit feedback. In the case of bandit feedback, the expected regret has an additional multiplicative term where reflects the number of learned parameters.
Cite
@article{arxiv.2212.11567,
title = {Learning Team Decisions},
author = {Olle Kjellqvist and Ather Gattami},
journal= {arXiv preprint arXiv:2212.11567},
year = {2022}
}
Comments
Accepted and presented at IEEE CDC 2022. A few typos have been corrected