Can Wikipedia Help Offline Reinforcement Learning? - Paper Explained
Offered By: Yannic Kilcher via YouTube
Course Description
Overview
Explore a comprehensive analysis of a research paper examining the potential of Wikipedia to enhance offline reinforcement learning. Delve into the innovative approach of treating reinforcement learning as sequence modeling, leveraging pre-trained language models to improve performance in control and game tasks. Discover how this method accelerates training by 3-6 times and achieves state-of-the-art results across various environments. Gain insights into the experimental findings, attention pattern analysis, and scaling properties of this novel technique. Understand the implications for bridging the gap between language modeling and reinforcement learning, opening new avenues for knowledge transfer between seemingly disparate domains.
Syllabus
- Intro
- Paper Overview
- Offline Reinforcement Learning as Sequence Modelling
- Input Embedding Alignment & other additions
- Main experimental results
- Analysis of the attention patterns across models
- More experimental results scaling properties, ablations, etc.
- Final thoughts
Taught by
Yannic Kilcher
Related Courses
Microsoft Bot Framework and Conversation as a PlatformMicrosoft via edX Unlocking the Power of OpenAI for Startups - Microsoft for Startups
Microsoft via YouTube Improving Customer Experiences with Speech to Text and Text to Speech
Microsoft via YouTube Stanford Seminar - Deep Learning in Speech Recognition
Stanford University via YouTube Select Topics in Python: Natural Language Processing
Codio via Coursera