75 - Reinforcement / Imitation Learning in NLP, with Hal Daumé III
NLP Highlights
00:00
The Role of Imitative Learning in Repetition Learning
In the traditional sort of structured prediction setting where the output that's structured, it's almost always the case that you have some intermediate signal about whether you're doing a good job. So in your own work on using imitation learning and machine translation, what would you do in order to approximate this loss? You don't know until the end. I mean, maybe you could craft heuristics or something, but by default, you don't know. But when you do have this access, you can often get much better algorithms.
Transcript
Play full episode
