Reinforcement Learning (COSC352)

Reinforcement Learning (RL) is an essential topic in the machine learning paradigm. Supervised and unsupervised learning approaches establish a decision function based on an example dataset. Reinforcement learning does not require an example dataset, but instead, determines an optimal policy based on a set of rules and its interaction with an environment. This unit will explored both model-based and model-free examples, including the Monte Carlo Decision Process (MDP), Q-Learning, Temporal Difference (TD) and Dynamic Programming. Later we will explored both conventional and deep learning techniques used in Reinforcement Learning. Upon completion of this unit students will be able to adapt a range of tools to stochastic problems in machine learning with an understanding of the taxonomy and while working with practical examples.