A smoothing approach for composite conditional gradient with nonsmooth loss

Federico Pierucci 1 Zaid Harchaoui 1 Jérôme Malick 2
1 LEAR - Learning and recognition in vision
Inria Grenoble - Rhône-Alpes, LJK - Laboratoire Jean Kuntzmann, INPG - Institut National Polytechnique de Grenoble
2 BIPOP - Modelling, Simulation, Control and Optimization of Non-Smooth Dynamical Systems
Inria Grenoble - Rhône-Alpes, LJK - Laboratoire Jean Kuntzmann, INPG - Institut National Polytechnique de Grenoble
Abstract : We consider learning problems where the non-smoothness lies both in the convex empirical risk and in the regularization penalty. Examples of such problems include learning with nonsmooth loss functions and atomic decomposition regularization penalty. Such doubly nonsmooth learning problems prevent the use of recently proposed composite conditional gradient algorithms for training, which are particularly attractive for large-scale applications. Indeed, they rely on the assumption that the empirical risk part of the objective is smooth. We propose a composite conditional gradient algorithm with smoothing to tackle such learning problems. We set up a framework allowing to systematically design parametrized smooth surrogates of nonsmooth loss functions. We then propose a smoothed composite conditional gradient algorithm, for which we prove theoretical guarantees on the accuracy. We present promising experimental results on collaborative filtering tasks.
Complete list of metadatas

Cited literature [20 references]  Display  Hide  Download

https://hal.inria.fr/hal-01096630
Contributor : Thoth Team <>
Submitted on : Thursday, January 15, 2015 - 12:30:45 PM
Last modification on : Monday, April 1, 2019 - 4:46:04 PM
Long-term archiving on : Thursday, September 10, 2015 - 11:41:05 PM

File

RR-8662.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01096630, version 1

Citation

Federico Pierucci, Zaid Harchaoui, Jérôme Malick. A smoothing approach for composite conditional gradient with nonsmooth loss. [Research Report] RR-8662, INRIA Grenoble. 2014. ⟨hal-01096630⟩

Share

Metrics

Record views

1231

Files downloads

570