Deep Reinforcement Learning

3 credits · Graduate

Reinforcement Learning (RL), as a branch of machine learning focuses on goal-directed learning from interactions and sequential decision making in complex, dynamic, and uncertain environments. Potential applications of RL are diverse and include applications in areas of robotics, transportation, game playing, healthcare, psychology, neuroscience, industrial process control, and business management, to name a few. This course focuses on deep RL, which uses deep learning (DL) for function approximation in RL methods. DL uses artificial neural networks to extract relevant features from unstructured data, enabling deep RL agents to make decisions in complex high-dimensional environments. Students will learn about the recent advances in the field of deep RL, including a wide range of deep RL methods, their applicability to different scenarios, their strengths, and their limitations. CALENDAR NOTES: Prerequisite knowledge should include basic probability and statistics; basic calculus and linear algebra; foundations of machine learning; familiarity with deep learning including but not limited to concepts such as backpropagaton, convolutional neural networks, and recurrent neural networks; and proficiency in Python EXCLUSIONS: CSCI 4250