We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Statistics > Machine Learning

Title: Tractable Density Estimation on Learned Manifolds with Conformal Embedding Flows

Abstract: Normalizing flows are generative models that provide tractable density estimation via an invertible transformation from a simple base distribution to a complex target distribution. However, this technique cannot directly model data supported on an unknown low-dimensional manifold, a common occurrence in real-world domains such as image data. Recent attempts to remedy this limitation have introduced geometric complications that defeat a central benefit of normalizing flows: exact density estimation. We recover this benefit with Conformal Embedding Flows, a framework for designing flows that learn manifolds with tractable densities. We argue that composing a standard flow with a trainable conformal embedding is the most natural way to model manifold-supported data. To this end, we present a series of conformal building blocks and apply them in experiments with synthetic and real-world data to demonstrate that flows can model manifold-supported distributions without sacrificing tractable likelihoods.
Comments: NeurIPS 2021 Camera-Ready. Code: this https URL
Subjects: Machine Learning (stat.ML); Machine Learning (cs.LG)
Cite as: arXiv:2106.05275 [stat.ML]
  (or arXiv:2106.05275v2 [stat.ML] for this version)

Submission history

From: Brendan Ross [view email]
[v1] Wed, 9 Jun 2021 18:00:00 GMT (5391kb,D)
[v2] Thu, 11 Nov 2021 19:06:15 GMT (8427kb,D)

Link back to: arXiv, form interface, contact.