[13] Adji Bousso Dieng - Deep Probabilistic Graphical Modeling
The Thesis Review
00:00
The Problem With Topic Models
The model is inspired by the LDA type topic models. It works very well on many different settings. You kind of specify the notion of topic through the generative process. And I guess it's like an empirical thing then that you actually look at what topics are learned and those kind of correspond to our intuitive idea of topics. Yes. Just looking at what is being shared, what that's supposed to represent when you discover it from data at a high level. That's what it is.
Transcript
Play full episode
