Reinforcement Learning

To realize the full potential of AI, autonomous systems must learn to make good decisions; reinforcement learning (RL) is a powerful paradigm for doing so. RL algorithms are applicable to a wide range of tasks, including robotics, game playing, consumer modeling, and healthcare. In this course, you will gain a solid introduction to the field of reinforcement learning. Through a combination of lectures and coding assignments, you will become well versed in the core approaches and challenges in the field, including generalization and exploration. You will also have a chance to explore the concept of deep reinforcement learning–an extremely promising new area that combines reinforcement learning with deep learning techniques.

What you will learn

• Key features of reinforcement learning
• Reinforcement learning algorithms
• Policy iteration, temporal difference learning and Q-learning
• Linear value approximation
• MDP, POMDP, bandit, batch offline and online reinforcement learning
• Open challenges and hot topics in reinforcement learning

Prerequisites

• Proficiency in Python: All coding assignments will be written in Python. You should be familiar with numpy and matplotlib, as well as basic shell commands (ssh, scp, ls, cd, rm, mv, cp, zip, etc.).
• Calculus and Linear Algebra: You should understand the following concepts from multivariable calculus and linear algebra: chain rule, gradients, matrix multiplication, matrix inverse.
• Probability: You should be familiar with basic probability distributions and be able to define the following concepts for both continuous and discrete random variables: Expectation, independence, probability distribution functions, and cumulative distribution functions.
• Foundations of Machine Learning (Recommended): Knowledge of basic machine learning and/or deep learning is helpful, but not required.

Notes

This professional online course, based on the Winter 2021 Stanford graduate course CS234, features:

• Lecture videos edited and segmented to focus on essential content
• Coding assignments in which you will apply course topics to real-life models
• Office hours and support from Stanford-affiliated Course Facilitators
• Cohort structure providing opportunities to network and collaborate with motivated learners from diverse locations and professional backgrounds