Explore a 32-minute video that delves into the concept of learning from passive data in reinforcement learning. Discover how an innovative algorithm based on successor features can leverage passive data to gain insights about the environment without direct interaction. Follow along as the video breaks down the key components, starting with an introduction to offline reinforcement learning and successor features. Examine the proposed algorithm in detail, analyze its results, and consider critical perspectives on its implementation. Gain valuable insights into this cutting-edge approach that aims to reduce the costs associated with environment interaction in reinforcement learning.
Overview
Syllabus
- Intro
- Offline-RL
- Successor Features
- Algorithm
- Results
- Criticisms & Thoughts
Taught by
Edan Meyer
Related Courses
-
Learning Fast with No Goals - VISR Explained
-
Fast Reinforcement Learning With Generalized Policy Updates - Paper Explained
-
Generalization and Robustness in Offline Reinforcement Learning
-
Decision Transformer - Reinforcement Learning via Sequence Modeling
-
Intermediate Data Manipulation and Machine Learning
-
Offline Reinforcement Learning and Model-Based Optimization
