Reinforcement learning with latent representations

 When working with complex high-dimensional objects, it is usually beneficial to translate them into compact low-dimensional forms. In my talk, I will discuss some successful approaches that learn meaningful representations for efficient reinforcement learning.