Offline RL frames next-best action as a constrained decision process
In a rideshare coupon study, the authors report 19% higher lifetime value than the behavior policy.
Stories per day · 30 days
2 RSS
Relatedcausal-inference1next-best-action1offline-rl1primal-dual1reinforcement-learning1rideshare1soft-actor-critic1
2 stories, newest first · one card per story
Stories tagged #constrained-mdp
Parametric resonance can destabilize primal-dual learning in constrained MDPs
Mismatched constraint estimates drive resonance at twice the natural frequency; a shared estimate removes it but adds bias.