Temporal Predictive Coding For Model-Based Planning In Latent Space

           Tung Nguyen * 1 Rui Shu * 2 Tuan Pham * 1 Hung Bui 1 Stefano Ermon 2

            Abstract                 ever, it is well known that performing RL directly in the
                                 high-dimensional observation space is sample-inefficient
   High-dimensional observations are a major chal-       and may require a large amount of training data (Lake et al.,
   lenge in the application of model-based rei ...