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

Download:

Current browse context:

math.ST

Change to browse by:

References & Citations

Bookmark

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

Mathematics > Statistics Theory

Title: A Consistent Extension of Discrete Optimal Transport Maps for Machine Learning Applications

Abstract: Optimal transport maps define a one-to-one correspondence between probability distributions, and as such have grown popular for machine learning applications. However, these maps are generally defined on empirical observations and cannot be generalized to new samples while preserving asymptotic properties. We extend a novel method to learn a consistent estimator of a continuous optimal transport map from two empirical distributions. The consequences of this work are two-fold: first, it enables to extend the transport plan to new observations without computing again the discrete optimal transport map; second, it provides statistical guarantees to machine learning applications of optimal transport. We illustrate the strength of this approach by deriving a consistent framework for transport-based counterfactual explanations in fairness.
Subjects: Statistics Theory (math.ST)
Cite as: arXiv:2102.08644 [math.ST]
  (or arXiv:2102.08644v1 [math.ST] for this version)

Submission history

From: Alberto Gonzalez-Sanz [view email]
[v1] Wed, 17 Feb 2021 09:22:00 GMT (26kb)

Link back to: arXiv, form interface, contact.