Graph Neural Thompson Sampling

arXiv:2406.10686v1 Announce Type: new
Abstract: We consider an online decision-making problem with a reward function defined over graph-structured data. We formally formulate the problem as an instance of graph action bandit. We then propose \texttt{GNN-TS}, a Graph Neural Network (GNN) powered Thompson Sampling (TS) algorithm which employs a GNN approximator for estimating the mean reward function and the graph neural tangent features for uncertainty estimation. We prove that, under certain boundness assumptions on the reward function, GNN-TS achieves a state-of-the-art …

abstract action algorithm arxiv cs.ai cs.lg data decision function gnn graph graph neural network instance making mean network neural network problem sampling stat.ml structured data type