A short introduction to model-based Reinforcement Learning

Model-based RL learns a deterministic or probabilistic forward model of the dynamics of an environment for action planning or direct policy optimisation. This knowledge enables agents to learn with up to orders of magnitude fewer samples than their model-free counterparts. The cost however is lower maximal performance, due to model misspecification.

In this seminar we go through the basic approaches of model-based RL following Sergey Levine’s excellent lecture on the subject.

In this series →

Data Efficiency

Reinforcement Learning

Software: Tianshou: An elegant deep reinforcement … Trainings: Safe and efficient deep reinforcement … Trainings: Classical and modern methods in planning … Seminar: Episode-based RL with Movement Primitive Seminar: Double Gumbel Q-learning Pills: Hyperparameters in Reinforcement … Pills: Exploiting past success in Off-Policy … Pills: Jump-Start Reinforcement Learning Pills: Implicit Q Learning Pills: Advantage-Induced Policy Alignment Pills: Critic Regularized Regression Pills: Deep Reinforcement Learning at the Edge … Pills: Planning with Diffusion for Flexible … Pills: Resolving Training Biases via … Blog: Natural, Trust Region and Proximal … Seminar: Problems and solutions in offline … Seminar: An introduction to Unsupervised Neural … Seminar: Dealing with imbalanced classes Seminar: Neural Network Dynamics for Model-Based … Seminar: An introduction to data distillation