Explore cutting-edge deep learning methods with Professor Pieter Abbeel in this comprehensive 1 hour 43 minute lecture from Full Stack Deep Learning. Dive into advanced topics including few-shot learning, reinforcement learning, imitation learning, domain randomization, architecture search, and unsupervised learning. Discover how these techniques are applied in real-world scenarios, from game-playing to robotics. Learn about the challenges and success stories in each area, and gain insights into bridging the gap between research and practical applications. Understand the importance of computing power in achieving better results and explore strategies for staying up-to-date with the latest developments in the field. Gain valuable knowledge on reading academic papers, forming study groups, and utilizing resources to keep pace with the rapidly evolving world of deep learning.
Pieter Abbeel on Research Directions - Full Stack Deep Learning - November 2019
Overview
Syllabus
Intro
Supervised Learning
Model Agnostic Metalearning
Data Sets
References
Optimization
Reptile vs Mammal
Reinforcement Learning
What is reinforcement learning
Learning to play video games
Learning to control robots
How good is the learning
How will we bridge this
Under the hood
Parameterize
Recurrent Neural Network
Bandits
Navigation
Metalearning
Metaworld
Algorithms
Diabetes
Meta learning
imitation learning
Taught by
The Full Stack
