UCB CS285: Deep Reinforcement Learning
Replication of classical reinforcement learning algorithms
UCB CS285: Deep Reinforcement Learning
Course Info
Amazing open source course to learn ideas about modern reinforcement learning.
Lecturer: Sergey Levine, Personal Website and his outstanding Berkeley RAIL lab
Course Website
Lectures
Deep RL 2023 class is a well developed version. The lectures and videos are more challenging than previous versions.
Assignments
Course assignment are separated into theory and application parts. I’ve completed hw1-3 model-free models and algorithms on Mujoco dataset.
- Homework 1: Imitation Learning: Requirements, Solution.
Replicated Behavioral Cloning and DAGGAR algorithm
- Homework 2: Policy Gradients: Requirements, Solution.
Replicated Policy Gradients, Baseline, Generalized Advantage Estimation(GAE)
- Homework 3: Q-learning and Actor-Critic Algorithms: Requirements, Solution
Replicated Deep Q-Networks(DQN), Double Q-Learning(DDQN), Actor-Critic, Reparameterize, and Soft Actor-Critic(SAC)
This post is licensed under CC BY 4.0 by the author.