Cut Pursuit: fast algorithms to learn piecewise constant functions

Loic Landrieu; Guillaume Obozinski

Conference papers

Cut Pursuit: fast algorithms to learn piecewise constant functions

Loic Landrieu ^{1, 2, 3, 4} Guillaume Obozinski ^{1, 2, 3}

CSTB - Centre Scientifique et Technique du Bâtiment, ENPC - École des Ponts ParisTech, LIGM - Laboratoire d'Informatique Gaspard-Monge

3 LIGM - Laboratoire d'Informatique Gaspard-Monge

4 SIERRA - Statistical Machine Learning and Parsimony

DI-ENS - Département d'informatique de l'École normale supérieure, CNRS - Centre National de la Recherche Scientifique, Inria de Paris

Abstract : We propose working-set/greedy algorithms to efficiently find the solutions to convex optimization problems penalized respectively by the total variation and the Mumford Shah boundary size. Our algorithms exploit the piecewise constant structure of the level-sets of the solutions by recursively splitting them using graph cuts. We obtain significant speed up on images that can be approximated with few level-sets compared to state-of-the-art algorithms .

Keywords : Mumford-Shah minimal partition optimization working-set greedy algorithm

Document type :

Conference papers

Domain :

Statistics [stat] / Machine Learning [stat.ML]

Complete list of metadata

Cited literature [4 references] Display Hide Download

https://hal.archives-ouvertes.fr/hal-01306786
Contributor : Loic Landrieu <>
Submitted on : Monday, April 25, 2016 - 3:17:40 PM
Last modification on : Tuesday, May 4, 2021 - 2:06:02 PM
Long-term archiving on: : Tuesday, November 15, 2016 - 11:45:10 AM

aistat2016_landrieu.pdf

Files produced by the author(s)

aistat2016-landrieu-supp.pdf Files produced by the author(s)

Cut Pursuit: fast algorithms to learn piecewise constant functions

Files

Identifiers

Collections

Citation

Export

Metrics

586

1005

Cut Pursuit: fast algorithms to learn piecewise constant functions

Files

Identifiers

Collections

Citation

Export

Share

Metrics

586

1005