Statistics > Machine Learning
[Submitted on 21 May 2015 (v1), last revised 14 Jun 2016 (this version, v6)]
Title:Variational Inference with Normalizing Flows
View PDFAbstract:The choice of approximate posterior distribution is one of the core problems in variational inference. Most applications of variational inference employ simple families of posterior approximations in order to allow for efficient inference, focusing on mean-field or other simple structured approximations. This restriction has a significant impact on the quality of inferences made using variational methods. We introduce a new approach for specifying flexible, arbitrarily complex and scalable approximate posterior distributions. Our approximations are distributions constructed through a normalizing flow, whereby a simple initial density is transformed into a more complex one by applying a sequence of invertible transformations until a desired level of complexity is attained. We use this view of normalizing flows to develop categories of finite and infinitesimal flows and provide a unified view of approaches for constructing rich posterior approximations. We demonstrate that the theoretical advantages of having posteriors that better match the true posterior, combined with the scalability of amortized variational approaches, provides a clear improvement in performance and applicability of variational inference.
Submission history
From: Danilo Jimenez Rezende [view email][v1] Thu, 21 May 2015 15:36:37 UTC (5,463 KB)
[v2] Fri, 22 May 2015 09:13:28 UTC (5,463 KB)
[v3] Tue, 26 May 2015 15:46:33 UTC (5,485 KB)
[v4] Mon, 22 Jun 2015 18:36:32 UTC (5,485 KB)
[v5] Mon, 13 Jun 2016 08:46:44 UTC (5,485 KB)
[v6] Tue, 14 Jun 2016 09:01:36 UTC (5,485 KB)
Current browse context:
stat.ME
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
Connected Papers (What is Connected Papers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.