Method
Live-Online
Term
SUMMER
Units
3.0 QUARTER UNITS
Estimated Cost
$980

Skills you will gain

  • Master MDP Formulation: Understand and implement Markov decision processes, value functions, policies, and the Bellman equation for RL problem-solving.
  • Leverage Modern DRL Algorithms: Apply deep Q-learning (DQN, DDQN), policy gradient methods (A2C, A3C, PPO), GRPO for reasoning, and inverse reinforcement learning to complex tasks.
  • Hands-on RL Coding: Use Python 3, Gymnasium, tf.keras, and popular DRL libraries such as Google Dopamine, Keras-RL, Hugging Face TRL, and Facebook Horizon to build RL models.
  • Enhance RL with Deep Learning: Explore how deep neural networks improve reinforcement learning techniques and enable human-like decision-making.
  • Apply RLHF to Real-World Systems: Implement Reinforcement Learning from Human Feedback and PPO to understand how modern LLMs like GPT are trained — and build a lightweight ChatGPT-like system.

Course Description

Discover how reinforcement learning from human feedback (RLHF) powers ChatGPT-like systems. This advanced course covers deep reinforcement learning fundamentals through hands-on simulations like CartPole and MuJoCo, exploring algorithms behind game-playing AI, self-driving cars, and robotics. Code in Python and Gymnasium as you build and train your own DRL agents.

Prerequisites / Skills Needed

Prerequisites:

  • AISV.X401: Deep Learning and Artificial Intelligence
Syllabus Library
Currently no classes scheduled. Would you like to be notified when a class is available?

This course applies to these programs:

Demo