Rethinking Model-based, Policy-based, and Value-based Reinforcement Learning via the Lens of Representation Complexity

Guhao Feng, Han Zhong

Reinforcement Learning (RL) encompasses diverse paradigms, including model-based RL, policy-based RL, and value-based RL, each tailored to approximate the model, optimal policy, an…