Master reinforcement learning concepts. Implement a complete RL solution and understand how to leverage AI tools to solve real-world problems.
Suggested by: Coursera (What is Coursera?)
No prior knowledge required
No unnecessary risks
Reinforcement Learning specialization includes 4 courses that explore the power of adaptive learning systems and artificial intelligence (AI). To harness the full potential of artificial intelligence, adaptive learning systems are needed. You will learn how reinforcement learning (RL) solutions help solve real-world problems through trial-and-error interaction, by implementing a complete RL solution from start to finish.
By the end of the internship, learners will understand the fundamentals of many of the modern technologies in artificial intelligence (AI) and will be ready to move on to more advanced courses or apply AI ideation tools to real-world problems. The content will focus on “small-scale” problems to understand the fundamentals of reinforcement learning, while learning from world-renowned experts from the University of Alberta, Faculty of Science.
Through programming tasks and quizzes, students:
In this course, you will learn about several algorithms that can learn near-optimal policies based on interaction with the environment – learning from the agent’s personal experience. Learning from practical experience is impressive because it does not require prior knowledge of the dynamics of the environment, but can still achieve optimal behavior. We will discuss the simple but powerful Monte Carlo methods, and time-difference learning methods including Q-learning. We will conclude the course by exploring how we can combine the two worlds: algorithms that can combine model-based planning (similar to dynamic programming) and time-difference updates to dramatically speed up learning.
In this course, you will learn how to solve problems with large, high-dimensional, and potentially infinite state spaces. You will see that evaluating value functions can be presented as a supervised learning problem—function optimization—that allows you to build agents that carefully balance generalization and differentiation to maximize reward. We will begin this journey by exploring how policy evaluation or prediction methods such as Monte Carlo and TD can be extended to define function optimization. You will learn about feature building techniques for RL and learning representations using neural networks and recurrence. We will conclude this course with an in-depth look at policy gradient methods; a way to learn policies directly without learning a value function. In this course, you will solve two continuous-state control tasks and explore the benefits of policy gradient methods in a continuous-state environment. Prerequisites: This course builds heavily on the foundations of courses 1 and 2, and students should complete these before starting this course. Students should also be comfortable with probability and expectations, basic linear algebra, basic calculus, Python 3.0 (at least one year), and implementing algorithms from pseudocode.
In this final course, you will combine your knowledge from courses 1, 2, and 3 to implement a complete RL solution to a problem. This capstone will allow you to see how each component—problem formulation, algorithm selection, parameter selection, and representation design—fits together into a complete solution, and how to make appropriate choices when implementing RL in the real world. This project will require you to implement both the environment for your problem’s stimulation and a control agent with neural network function optimization. In addition, you will conduct a scientific study of your learning system to develop your ability to evaluate the robustness of RL agents. To use RL in the real world, it is critical to (a) properly formulate the problem as a Markov decision process, (b) select the appropriate algorithms, (c) identify which choices in your implementation will have a large impact on performance, and (d) validate the expected behavior of your algorithms. This capstone is useful for anyone who plans to use RL to solve real-world problems. To succeed in this course, you will need to have completed courses 1, 2, and 3 of this specialization or their equivalent.
