InfoCoBuild

CS224R - Deep Reinforcement Learning

CS224R - Deep Reinforcement Learning (Spring 2025, Stanford Univ.). Instructor: Prof. Chelsea Finn. Decision-making is central to modern AI systems - from robots and autonomous vehicles to chip design and large language models. Capable AI systems must act and make decisions, not just predict. Deep reinforcement learning (RL) enables this by learning from consequences and feedback, using deep networks to handle high-dimensional observations and complex dynamics.

In this course, you will study practical algorithms for deep RL and how neural networks represent policies, value functions, and world models. You will build methods that learn directly from experience and gain hands-on experience with training, fine-tuning, and evaluating agents on real tasks. (from Stanford Online)

Lecture 15 - Hierarchical RL and IL


Go to the Course Home or watch other lectures:

Lecture 01 - Class Intro
Lecture 02 - Limitation Learning
Lecture 03 - Policy Gradients
Lecture 04 - Actor-Critic Methods
Lecture 05 - Off-Policy Actor-Critic
Lecture 06 - Q-Learning
Lecture 07 - Offline RL
Lecture 08 - Reward Learning
Lecture 09 - RL for LLMs
Lecture 10 - RL for LLM Reasoning
Lecture 11 - Model-Based RL
Lecture 12 - Multi-Task RL
Lecture 13 - Meta RL
Lecture 14 - Exploration
Lecture 15 - Hierarchical RL and IL
Lecture 16 - RL for Robots
Lecture 17 - Advancing Robot Intelligence
Lecture 18 - Frontiers
Lecture 19 - Tutorial Session: Review of Q-Learning