Description
| Title | Normalizing flows for discrete data |
| Abstract | In this talk, I will give an introduction to normalising flows. Normalising flows allow us to learn densities through the use of invertible neural networks with the goal of being able to model complicated dependency structures and multiple modes. This has applications in variational inference and generative modelling. After the introduction, I will present three recent works that focus on normalising flows for discrete data as opposed to real-valued data and their application to lossless compression. |
Other presentations by Rianne van den Berg (Research Scientist, Google Brain)
| Date | Title |
|---|---|
| 09 March 2020 | Normalizing flows for discrete data |