Deep Learning For Natural Language Processing
EE-608
Video of talk "Bayesian Language Understanding with Nonparametric Variational Transformers"
This page is part of the content downloaded from Video of talk "Bayesian Language Understanding with Nonparametric Variational Transformers" on Sunday, 25 January 2026, 00:15. Note that some content and any files larger than 50 MB are not downloaded.
Description
Invited talk given at the CALCULUS Symposium, at KU Leuven in 2024.
We present our work on reinterpreting the latent representations of Transformers as nonparametric mixture distributions, and training a variational Bayesian version of Transformers on natural language. We argue that the empirical success of Transformers is evidence that natural language understanding is nonparametric variational Bayesian inference of mixture distributions.