
Episode 07: Yujia Huang, Caltech, on neuro-inspired generative models
Generally Intelligent
00:00
How to Use Generative Model to Help Discriminative Tasks
If you have a consistent label and consistent input, if these work consistent with each other, you shouldn't change your label or the impression in your brain. It's much harder to model a high dimensional distribution like images than just doing an image classification problem. Sometimes, you might not want to predict all the aspects of the image. So that's why in my paper, actually, I choose which layer to add feedback to.
Transcript
Play full episode