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

Download:

Current browse context:

stat.ML

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

Statistics > Machine Learning

Title: Beyond Perturbation Stability: LP Recovery Guarantees for MAP Inference on Noisy Stable Instances

Abstract: Several works have shown that perturbation stable instances of the MAP inference problem in Potts models can be solved exactly using a natural linear programming (LP) relaxation. However, most of these works give few (or no) guarantees for the LP solutions on instances that do not satisfy the relatively strict perturbation stability definitions. In this work, we go beyond these stability results by showing that the LP approximately recovers the MAP solution of a stable instance even after the instance is corrupted by noise. This "noisy stable" model realistically fits with practical MAP inference problems: we design an algorithm for finding "close" stable instances, and show that several real-world instances from computer vision have nearby instances that are perturbation stable. These results suggest a new theoretical explanation for the excellent performance of this LP relaxation in practice.
Comments: 25 pages, 2 figures, 2 tables. To appear in AISTATS 2021
Subjects: Machine Learning (stat.ML); Machine Learning (cs.LG)
Cite as: arXiv:2103.00034 [stat.ML]
  (or arXiv:2103.00034v1 [stat.ML] for this version)

Submission history

From: Hunter Lang [view email]
[v1] Fri, 26 Feb 2021 20:01:44 GMT (136kb,D)

Link back to: arXiv, form interface, contact.