A Complete Reinforcement Learning System (Capstone)

Overview

In the comprehensive capstone course, "A Complete Reinforcement Learning System," offered by the University of Alberta on Coursera, learners have a unique opportunity to synthesize their acumen gathered across the initial three courses into devising a full-fledged reinforcement learning (RL) solution tailored to a specific problem. This pivotal project empowers participants to not only combine the integral elements of problem scoping, algorithmic selection, parameter tuning, and representation crafting into a cohesive RL application but also enables them to make informed decisions when applying RL strategies in real-world scenarios. As part of the capstone, learners will undertake the development of an environment to stimulate their chosen problem and a control agent leveraging Neural Network for function approximation. Furthermore, the course includes conducting a scientific analysis to evaluate the resilience of RL agents, emphasizing the importance of accurately framing issues as Markov Decision Processes (MDP), choosing suitable algorithms, understanding the impact of implementation decisions on performance, and ensuring the algorithms behave as anticipated. This capstone stands as an invaluable asset for anyone looking to apply RL to solve practical challenges.

Prerequisites for this course include the successful completion of the first three courses in the Specialization or their equivalences. Upon concluding this course, participants will be proficient in implementing a comprehensive RL solution—beginning from problem formulation, through to selecting and implementing the right algorithms, and conducting empirical studies to affirm the solution's effectiveness. Discover the realms of Artificial Intelligence and Reinforcement Learning through this capstone offering, categorized under Artificial Intelligence Courses and Reinforcement Learning Courses.

Syllabus

Taught by

Martha White and Adam White

Tags

canada