2 problems

0 votes0 replies0 views
The wider-regime conjecture for minimax lower bounds in off-policy evaluation
Wider-regime conjecture. A similar lower bound should hold in the wider regime $\nu_0^{-1}\lesssim n\lesssim|\mathcal{S}|$ .
Conjectureopenminimax-lower-bounds off-policy-evaluation reinforcement-learningarXiv posted 2 days ago
0 votes0 replies1 view
Robustness of an $L^1$ -based Bellman-error estimator under heavy-tailed rewards
Let $Q$ be a $Q$ -function and let $r$ be the reward function. An estimator can be constructed by minimizing an empirical approximation of the $L^1$ -norm of the Bellman error (resid…
Conjectureopenoff-policy-evaluation reinforcement-learning robust-statisticsarXiv posted 2 days ago

The wider-regime conjecture for minimax lower bounds in off-policy evaluation