VQ-Flows: Vector quantized local normalizing flows

Sahil Sidheekh, Chris B. Dock, Tushar Jain, Radu Balan, Maneesh K. Singh
Proceedings of the Thirty-Eighth Conference on Uncertainty in Artificial Intelligence, PMLR 180:1835-1845, 2022.

Abstract

Normalizing flows provide an elegant approach to generative modeling that allows for efficient sampling and exact density evaluation of unknown data distributions. However, current techniques have significant limitations in their expressivity when the data distribution is supported on a low-dimensional manifold or has a non-trivial topology. We introduce a novel statistical framework for learning a mixture of local normalizing flows as “chart maps” over the data manifold. Our framework augments the expressivity of recent approaches while preserving the signature property of normalizing flows, that they admit exact density evaluation. We learn a suitable atlas of charts for the data manifold via a vector quantized auto-encoder (VQ-AE) and the distributions over them using a conditional flow. We validate experimentally that our probabilistic framework enables existing approaches to better model data distributions over complex manifolds.

Cite this Paper

BibTeX
@InProceedings{pmlr-v180-sidheekh22a, title = {VQ-Flows: Vector quantized local normalizing flows}, author = {Sidheekh, Sahil and Dock, Chris B. and Jain, Tushar and Balan, Radu and Singh, Maneesh K.}, booktitle = {Proceedings of the Thirty-Eighth Conference on Uncertainty in Artificial Intelligence}, pages = {1835--1845}, year = {2022}, editor = {Cussens, James and Zhang, Kun}, volume = {180}, series = {Proceedings of Machine Learning Research}, month = {01--05 Aug}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v180/sidheekh22a/sidheekh22a.pdf}, url = {https://proceedings.mlr.press/v180/sidheekh22a.html}, abstract = {Normalizing flows provide an elegant approach to generative modeling that allows for efficient sampling and exact density evaluation of unknown data distributions. However, current techniques have significant limitations in their expressivity when the data distribution is supported on a low-dimensional manifold or has a non-trivial topology. We introduce a novel statistical framework for learning a mixture of local normalizing flows as “chart maps” over the data manifold. Our framework augments the expressivity of recent approaches while preserving the signature property of normalizing flows, that they admit exact density evaluation. We learn a suitable atlas of charts for the data manifold via a vector quantized auto-encoder (VQ-AE) and the distributions over them using a conditional flow. We validate experimentally that our probabilistic framework enables existing approaches to better model data distributions over complex manifolds.} }
Endnote
%0 Conference Paper %T VQ-Flows: Vector quantized local normalizing flows %A Sahil Sidheekh %A Chris B. Dock %A Tushar Jain %A Radu Balan %A Maneesh K. Singh %B Proceedings of the Thirty-Eighth Conference on Uncertainty in Artificial Intelligence %C Proceedings of Machine Learning Research %D 2022 %E James Cussens %E Kun Zhang %F pmlr-v180-sidheekh22a %I PMLR %P 1835--1845 %U https://proceedings.mlr.press/v180/sidheekh22a.html %V 180 %X Normalizing flows provide an elegant approach to generative modeling that allows for efficient sampling and exact density evaluation of unknown data distributions. However, current techniques have significant limitations in their expressivity when the data distribution is supported on a low-dimensional manifold or has a non-trivial topology. We introduce a novel statistical framework for learning a mixture of local normalizing flows as “chart maps” over the data manifold. Our framework augments the expressivity of recent approaches while preserving the signature property of normalizing flows, that they admit exact density evaluation. We learn a suitable atlas of charts for the data manifold via a vector quantized auto-encoder (VQ-AE) and the distributions over them using a conditional flow. We validate experimentally that our probabilistic framework enables existing approaches to better model data distributions over complex manifolds.
APA
Sidheekh, S., Dock, C.B., Jain, T., Balan, R. & Singh, M.K.. (2022). VQ-Flows: Vector quantized local normalizing flows. Proceedings of the Thirty-Eighth Conference on Uncertainty in Artificial Intelligence, in Proceedings of Machine Learning Research 180:1835-1845 Available from https://proceedings.mlr.press/v180/sidheekh22a.html.

Related Material