Finite-Time Convergence of Single-Trajectory Chi-Square Robust Q-Learning With Linear Function Approximation
This work addresses obstacles through a variational reformulation of the robust Bellman target and a blockwise frozen-target scheme, and establishes a finite-time error bound relative to the optimal robust Q-function for every $\gamma\in(0,1)$.
Saptarshi Mandal, Y. Murthy, R. Srikant
· 0 citations