Learning Parsimonious Dynamics for Generalization in Reinforcement Learning
This addresses generalization issues in reinforcement learning for agents navigating novel environments, though it appears incremental as it builds on existing latent learning principles.
The paper tackles the problem of poor generalization in model-based reinforcement learning by proposing a model that learns parsimonious latent dynamics, demonstrating its utility in policy learning and planning tasks.
Humans are skillful navigators: We aptly maneuver through new places, realize when we are back at a location we have seen before, and can even conceive of shortcuts that go through parts of our environments we have never visited. Current methods in model-based reinforcement learning on the other hand struggle with generalizing about environment dynamics out of the training distribution. We argue that two principles can help bridge this gap: latent learning and parsimonious dynamics. Humans tend to think about environment dynamics in simple terms -- we reason about trajectories not in reference to what we expect to see along a path, but rather in an abstract latent space, containing information about the places' spatial coordinates. Moreover, we assume that moving around in novel parts of our environment works the same way as in parts we are familiar with. These two principles work together in tandem: it is in the latent space that the dynamics show parsimonious characteristics. We develop a model that learns such parsimonious dynamics. Using a variational objective, our model is trained to reconstruct experienced transitions in a latent space using locally linear transformations, while encouraged to invoke as few distinct transformations as possible. Using our framework, we demonstrate the utility of learning parsimonious latent dynamics models in a range of policy learning and planning tasks.