[10] Chelsea Finn - Learning to Learn with Gradients
The Thesis Review
00:00
The Motivation Behind Meta Learning
We wanted to build systems that can somehow leverage previous experience when learning new tasks. If we want robots to be broadly useful we want them to do many different skills, many different tasks. And it seemed like there wasn't any existing algorithm that would allow you to take previous experience and use that to learn a new skill extremely quickly. So that was some of the motivation behind meta learning.
Transcript
Play full episode
