Improved Model-based Reinforcement Learning with Smooth Kernels

ArXi:2605.07218v1 Announce Type: new For continuous state-action space scenarios, classical reinforcement learning (RL) theory predominantly focuses on low-rank Marko decision processes (MDPs), which provide sample-efficient guarantees at the expense of restrictive structural assumptions. Kernel smoothing model-based approaches offer a promising alternative paradigm that instead leverages the smoothness of the MDP and employs non-parametric kernel smoothing estimates of transition dynamics.