Reinforcement Learning with Deep Energy-Based Policies
This work addresses the challenge of applying energy-based policies to continuous control tasks, offering potential benefits for robotics and reinforcement learning, though it appears incremental as it builds on existing methods like maximum entropy policies and Stein variational gradient descent.
The authors tackled the problem of learning expressive energy-based policies for continuous states and actions, which was previously feasible only in tabular domains, by proposing soft Q-learning and using amortized Stein variational gradient descent, resulting in improved exploration and skill transferability in simulated robot experiments.
We propose a method for learning expressive energy-based policies for continuous states and actions, which has been feasible only in tabular domains before. We apply our method to learning maximum entropy policies, resulting into a new algorithm, called soft Q-learning, that expresses the optimal policy via a Boltzmann distribution. We use the recently proposed amortized Stein variational gradient descent to learn a stochastic sampling network that approximates samples from this distribution. The benefits of the proposed algorithm include improved exploration and compositionality that allows transferring skills between tasks, which we confirm in simulated experiments with swimming and walking robots. We also draw a connection to actor-critic methods, which can be viewed performing approximate inference on the corresponding energy-based model.