Applying probabilistic models to reinforcement learning (RL) enables the uses of powerful optimisation tools such as variational inference in RL. 2 and 3. My interests span probabilisitic inference, stochastic processes, Bayesian inference, Bayesian nonparametrics, Bayesian reinforcement learning, approximate inference (variational and MCMC), statistical signal processing, and infromation theory. • We combine variational information optimisation and tools from deep learning to develop a scal- able algorithm for intrinsically-motivated reinforcement learning, demonstrating a new applica- tion of the variational theory for problems in reinforcement learning and decision making. Varia-tional inference introduces an approximate distribution q (2. test- In the process, dozens of new algorithms have been proposed for solving these problems with deep neural networks, speciﬁc of course to domain at hand. We then focus on the connection between the two frameworks in Sec. Probability Functional Descent: A Unifying Perspective on GANs, Variational Inference, and Reinforcement Learning 2019 This paper provides a unifying view of a wide range of problems of interest in machine learning by framing them as the minimization of functionals defined on the space of probability measures. A recent line of research casts ‘RL as inference’ and suggests a particular framework to generalize the RL problem as probabilistic inference. Research . Reinforcement Learning. We provide background on variational inference and reinforcement learning in Secs. Abstract: Reinforcement learning (RL) combines a control problem with statistical estimation: The system dynamics are not known to the agent, but can be learned through experience. The basic idea of reinforcement learning is that we want to train an model to make a sequence of predictions, where after each prediction, we recieve a reward, and enter into a new state. 2 Model-based Reinforcement Learning as Bayesian Inference In this section, we describe MBRL as a Bayesian inference problem using control as inference framework . Fig.2displays the graphical model for the formulation, with which an MBRL procedure can be re-written in a Bayesian fashion: (1. training-step) do inference of p( jD). 5. Applying probabilistic models to reinforcement learning (RL) has become an exciting direction of research owing to powerful optimisation tools such as variational inference … Deep learning now plays an important role in many domains, for example, in generative modeling, deep reinforcement learning, and variational inference. 11/03/2018 ∙ by Matthew Fellows, et al.