Neural Spline Flows

Conor Durkan; Artur Bekasovs; Iain Murray; George Papamakarios

Neural Spline Flows

Conor Durkan, Artur Bekasovs, Iain Murray, George Papamakarios

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

A normalizing flow models a complex probability density as an invertible transformation of a simple base density. Flows based on either coupling or autoregressive transforms both offer exact density evaluation and sampling, but rely on the parameterization of an easily invertible elementwise transformation, whose choice determines the flexibility of these models. Building upon recent work, we propose a fully-differentiable module based on monotonic rational-quadratic splines, which enhances the flexibility of both coupling and autoregressive transforms while retaining analytic invertibility. We demonstrate that neural spline flows improve density estimation, variational inference, and generative modeling of images.

Original language	English
Title of host publication	Advances in Neural Information Processing Systems 32 (NeurIPS 2019)
Publisher	Neural Information Processing Systems Foundation, Inc
Pages	7511-7522
Number of pages	12
Volume	32
Publication status	Published - 14 Dec 2019
Event	33rd Conference on Neural Information Processing Systems - Vancouver Convention Centre, Vancouver, Canada Duration: 8 Dec 2019 → 14 Dec 2019 https://neurips.cc/

Publication series

Name	Advances in Neural Information Processing Systems
Volume	32
ISSN (Electronic)	1049-5258

Conference

Conference	33rd Conference on Neural Information Processing Systems
Abbreviated title	NeurIPS 2019
Country/Territory	Canada
City	Vancouver
Period	8/12/19 → 14/12/19
Internet address	https://neurips.cc/

Access to Document

Neural Spline Flows_DURKAN_DOA03092019_AFVAccepted author manuscript, 4.78 MB

https://papers.nips.cc/paper/8969-neural-spline-flowsLicence: All Rights Reserved

Iain Murray
- i.murrayed.acuk
- School of Informatics - Personal Chair of Machine Learning and Inference
- Institute for Adaptive and Neural Computation
- Data Science and Artificial Intelligence
Person: Academic: Research Active

Cite this

@inproceedings{d1220ead849147929d687c8a42669907,

title = "Neural Spline Flows",

abstract = "A normalizing flow models a complex probability density as an invertible transformation of a simple base density. Flows based on either coupling or autoregressive transforms both offer exact density evaluation and sampling, but rely on the parameterization of an easily invertible elementwise transformation, whose choice determines the flexibility of these models. Building upon recent work, we propose a fully-differentiable module based on monotonic rational-quadratic splines, which enhances the flexibility of both coupling and autoregressive transforms while retaining analytic invertibility. We demonstrate that neural spline flows improve density estimation, variational inference, and generative modeling of images.",

author = "Conor Durkan and Artur Bekasovs and Iain Murray and George Papamakarios",

year = "2019",

month = dec,

day = "14",

language = "English",

volume = "32",

series = "Advances in Neural Information Processing Systems",

publisher = "Neural Information Processing Systems Foundation, Inc",

pages = "7511--7522",

booktitle = "Advances in Neural Information Processing Systems 32 (NeurIPS 2019)",

note = "33rd Conference on Neural Information Processing Systems, NeurIPS 2019 ; Conference date: 08-12-2019 Through 14-12-2019",

url = "https://neurips.cc/",

}

Durkan, C , Bekasovs, A , Murray, I & Papamakarios, G 2019, Neural Spline Flows. in Advances in Neural Information Processing Systems 32 (NeurIPS 2019). vol. 32, Advances in Neural Information Processing Systems, vol. 32, Neural Information Processing Systems Foundation, Inc, pp. 7511-7522, 33rd Conference on Neural Information Processing Systems, Vancouver, British Columbia, Canada, 8/12/19. <https://papers.nips.cc/paper/8969-neural-spline-flows>

Neural Spline Flows. / Durkan, Conor ; Bekasovs, Artur ; Murray, Iain et al.

Advances in Neural Information Processing Systems 32 (NeurIPS 2019). Vol. 32 Neural Information Processing Systems Foundation, Inc, 2019. p. 7511-7522 (Advances in Neural Information Processing Systems; Vol. 32).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - Neural Spline Flows

AU - Durkan, Conor

AU - Bekasovs, Artur

AU - Murray, Iain

AU - Papamakarios, George

PY - 2019/12/14

Y1 - 2019/12/14

N2 - A normalizing flow models a complex probability density as an invertible transformation of a simple base density. Flows based on either coupling or autoregressive transforms both offer exact density evaluation and sampling, but rely on the parameterization of an easily invertible elementwise transformation, whose choice determines the flexibility of these models. Building upon recent work, we propose a fully-differentiable module based on monotonic rational-quadratic splines, which enhances the flexibility of both coupling and autoregressive transforms while retaining analytic invertibility. We demonstrate that neural spline flows improve density estimation, variational inference, and generative modeling of images.

AB - A normalizing flow models a complex probability density as an invertible transformation of a simple base density. Flows based on either coupling or autoregressive transforms both offer exact density evaluation and sampling, but rely on the parameterization of an easily invertible elementwise transformation, whose choice determines the flexibility of these models. Building upon recent work, we propose a fully-differentiable module based on monotonic rational-quadratic splines, which enhances the flexibility of both coupling and autoregressive transforms while retaining analytic invertibility. We demonstrate that neural spline flows improve density estimation, variational inference, and generative modeling of images.

M3 - Conference contribution

VL - 32

T3 - Advances in Neural Information Processing Systems

SP - 7511

EP - 7522

BT - Advances in Neural Information Processing Systems 32 (NeurIPS 2019)

PB - Neural Information Processing Systems Foundation, Inc

T2 - 33rd Conference on Neural Information Processing Systems

Y2 - 8 December 2019 through 14 December 2019

ER -

Neural Spline Flows

Abstract

Publication series

Conference

Access to Document

Fingerprint

Profiles

Iain Murray

Cite this