Nonstationary Multi-Armed Bandit
Nonstationary Multi-Armed Bandit: a task in openthoughts-tblite-dev-apptainer-v1: OpenThoughts-TBLite dev (offline Apptainer, v1) (Harbor dataset). Play a terminal-based nonstationary multi-armed bandit game. Choose arms, receive noisy rewards with drifting distributions, and minimize regret over…
The task
Play a terminal-based nonstationary multi-armed bandit game. Choose arms, receive noisy rewards with drifting distributions, and minimize regret over a fixed horizon.