variational-inference

Model-based RL methods that learn latent-variable models instead of trying to predict dynamics models in the observed space. The learned world model then can be used in planning effectively rather than being less efficiently, for instance in visual-based tasks, generating images for future time steps and feed them back into the model to predict the next ones, which requires more computation. ...

An autoencoder that differs from others (with deterministic encoder) by mapping each input $\mathbf{x}$ to a distribution over the possible values of the latent representation $\mathbf{z}$ from which $\mathbf{x}$ could have been generated. ...