World Models
This work addresses efficient and unsupervised learning in reinforcement learning, with potential applications in robotics and gaming, though it is incremental in combining existing generative and reinforcement learning techniques.
The paper tackles the problem of building generative neural network models for reinforcement learning environments, achieving the ability to train agents inside hallucinated dreams and transfer policies back to actual environments.
We explore building generative neural network models of popular reinforcement learning environments. Our world model can be trained quickly in an unsupervised manner to learn a compressed spatial and temporal representation of the environment. By using features extracted from the world model as inputs to an agent, we can train a very compact and simple policy that can solve the required task. We can even train our agent entirely inside of its own hallucinated dream generated by its world model, and transfer this policy back into the actual environment. An interactive version of this paper is available at https://worldmodels.github.io/