---
_id: '2051'
abstract:
- lang: eng
text: We show that the usual score function for conditional Markov networks can
be written as the expectation over the scores of their spanning trees. We also
show that a small random sample of these output trees can attain a significant
fraction of the margin obtained by the complete graph and we provide conditions
under which we can perform tractable inference. The experimental results confirm
that practical learning is scalable to realistic datasets using this approach.
author:
- first_name: Mario
full_name: Marchand, Mario
last_name: Marchand
- first_name: Su
full_name: Hongyu, Su
last_name: Hongyu
- first_name: Emilie
full_name: Emilie Morvant
id: 4BAC2A72-F248-11E8-B48F-1D18A9856A87
last_name: Morvant
orcid: 0000-0002-8301-7240
- first_name: Juho
full_name: Rousu, Juho
last_name: Rousu
- first_name: John
full_name: Shawe-Taylor, John
last_name: Shawe Taylor
citation:
ama: 'Marchand M, Hongyu S, Morvant E, Rousu J, Shawe Taylor J. Multilabel structured
output learning with random spanning trees of max-margin Markov networks. In:
Neural Information Processing Systems; 2014.'
apa: 'Marchand, M., Hongyu, S., Morvant, E., Rousu, J., & Shawe Taylor, J. (2014).
Multilabel structured output learning with random spanning trees of max-margin
Markov networks. Presented at the NIPS: Neural Information Processing Systems,
Neural Information Processing Systems.'
chicago: Marchand, Mario, Su Hongyu, Emilie Morvant, Juho Rousu, and John Shawe
Taylor. “Multilabel Structured Output Learning with Random Spanning Trees of Max-Margin
Markov Networks.” Neural Information Processing Systems, 2014.
ieee: 'M. Marchand, S. Hongyu, E. Morvant, J. Rousu, and J. Shawe Taylor, “Multilabel
structured output learning with random spanning trees of max-margin Markov networks,”
presented at the NIPS: Neural Information Processing Systems, 2014.'
ista: 'Marchand M, Hongyu S, Morvant E, Rousu J, Shawe Taylor J. 2014. Multilabel
structured output learning with random spanning trees of max-margin Markov networks.
NIPS: Neural Information Processing Systems.'
mla: Marchand, Mario, et al. Multilabel Structured Output Learning with Random
Spanning Trees of Max-Margin Markov Networks. Neural Information Processing
Systems, 2014.
short: M. Marchand, S. Hongyu, E. Morvant, J. Rousu, J. Shawe Taylor, in:, Neural
Information Processing Systems, 2014.
conference:
name: 'NIPS: Neural Information Processing Systems'
date_created: 2018-12-11T11:55:26Z
date_published: 2014-01-01T00:00:00Z
date_updated: 2021-01-12T06:54:59Z
day: '01'
extern: 1
main_file_link:
- open_access: '1'
url: https://hal.archives-ouvertes.fr/hal-01065586
month: '01'
oa: 1
publication_status: published
publisher: Neural Information Processing Systems
publist_id: '4996'
quality_controlled: 0
status: public
title: Multilabel structured output learning with random spanning trees of max-margin
Markov networks
type: conference
year: '2014'
...
---
_id: '2057'
abstract:
- lang: eng
text: 'In the past few years, a lot of attention has been devoted to multimedia
indexing by fusing multimodal informations. Two kinds of fusion schemes are generally
considered: The early fusion and the late fusion. We focus on late classifier
fusion, where one combines the scores of each modality at the decision level.
To tackle this problem, we investigate a recent and elegant well-founded quadratic
program named MinCq coming from the machine learning PAC-Bayesian theory. MinCq
looks for the weighted combination, over a set of real-valued functions seen as
voters, leading to the lowest misclassification rate, while maximizing the voters’
diversity. We propose an extension of MinCq tailored to multimedia indexing. Our
method is based on an order-preserving pairwise loss adapted to ranking that allows
us to improve Mean Averaged Precision measure while taking into account the diversity
of the voters that we want to fuse. We provide evidence that this method is naturally
adapted to late fusion procedures and confirm the good behavior of our approach
on the challenging PASCAL VOC’07 benchmark.'
alternative_title:
- LNCS
arxiv: 1
author:
- first_name: Emilie
full_name: Morvant, Emilie
id: 4BAC2A72-F248-11E8-B48F-1D18A9856A87
last_name: Morvant
orcid: 0000-0002-8301-7240
- first_name: Amaury
full_name: Habrard, Amaury
last_name: Habrard
- first_name: Stéphane
full_name: Ayache, Stéphane
last_name: Ayache
citation:
ama: 'Morvant E, Habrard A, Ayache S. Majority vote of diverse classifiers for late
fusion. In: Lecture Notes in Computer Science (Including Subseries Lecture
Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol
8621. Springer; 2014:153-162. doi:10.1007/978-3-662-44415-3_16'
apa: 'Morvant, E., Habrard, A., & Ayache, S. (2014). Majority vote of diverse
classifiers for late fusion. In Lecture Notes in Computer Science (including
subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
(Vol. 8621, pp. 153–162). Joensuu, Finland: Springer. https://doi.org/10.1007/978-3-662-44415-3_16'
chicago: Morvant, Emilie, Amaury Habrard, and Stéphane Ayache. “Majority Vote of
Diverse Classifiers for Late Fusion.” In Lecture Notes in Computer Science
(Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes
in Bioinformatics), 8621:153–62. Springer, 2014. https://doi.org/10.1007/978-3-662-44415-3_16.
ieee: E. Morvant, A. Habrard, and S. Ayache, “Majority vote of diverse classifiers
for late fusion,” in Lecture Notes in Computer Science (including subseries
Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics),
Joensuu, Finland, 2014, vol. 8621, pp. 153–162.
ista: 'Morvant E, Habrard A, Ayache S. 2014. Majority vote of diverse classifiers
for late fusion. Lecture Notes in Computer Science (including subseries Lecture
Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). IAPR: International
Workshop on Structural, Syntactic, and Statistical Pattern Recognition, LNCS,
vol. 8621, 153–162.'
mla: Morvant, Emilie, et al. “Majority Vote of Diverse Classifiers for Late Fusion.”
Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial
Intelligence and Lecture Notes in Bioinformatics), vol. 8621, Springer, 2014,
pp. 153–62, doi:10.1007/978-3-662-44415-3_16.
short: E. Morvant, A. Habrard, S. Ayache, in:, Lecture Notes in Computer Science
(Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes
in Bioinformatics), Springer, 2014, pp. 153–162.
conference:
end_date: 2014-08-22
location: Joensuu, Finland
name: 'IAPR: International Workshop on Structural, Syntactic, and Statistical Pattern
Recognition'
start_date: 2014-08-20
date_created: 2018-12-11T11:55:28Z
date_published: 2014-01-01T00:00:00Z
date_updated: 2021-01-12T06:55:01Z
day: '01'
department:
- _id: ChLa
doi: 10.1007/978-3-662-44415-3_16
ec_funded: 1
external_id:
arxiv:
- '1404.7796'
intvolume: ' 8621'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: http://arxiv.org/abs/1404.7796
month: '01'
oa: 1
oa_version: Preprint
page: 153 - 162
project:
- _id: 2532554C-B435-11E9-9278-68D0E5697425
call_identifier: FP7
grant_number: '308036'
name: Lifelong Learning of Visual Scene Understanding
publication: Lecture Notes in Computer Science (including subseries Lecture Notes
in Artificial Intelligence and Lecture Notes in Bioinformatics)
publication_status: published
publisher: Springer
publist_id: '4989'
quality_controlled: '1'
scopus_import: 1
status: public
title: Majority vote of diverse classifiers for late fusion
type: conference
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
volume: 8621
year: '2014'
...
---
_id: '2165'
abstract:
- lang: eng
text: 'In machine learning, the domain adaptation problem arrives when the test
(tar-get) and the train (source) data are generated from different distributions. A
key applied issue is thus the design of algorithms able to generalize on a new
distribution, for which we have no label information. We focus on learning classification
models defined as a weighted majority vote over a set of real-valued functions.
In this context, Germain et al. (2013) have shown that a measure of disagreement
between these functions is crucial to control. The core of this measure is a theoretical
bound—the C-bound (Lacasse et al., 2007)—which involves the disagreement and leads
to a well performing majority vote learn-ing algorithm in usual non-adaptative
supervised setting: MinCq. In this work,we propose a framework to extend MinCq
to a domain adaptation scenario.This procedure takes advantage of the recent perturbed
variation divergence between distributions proposed by Harel and Mannor (2012). Justified
by a theoretical bound on the target risk of the vote, we provide to MinCq a
tar-get sample labeled thanks to a perturbed variation-based self-labeling focused
on the regions where the source and target marginals appear similar. We also
study the influence of our self-labeling, from which we deduce an original process
for tuning the hyperparameters. Finally, our framework called PV-MinCq shows very
promising results on a rotation and translation synthetic problem.'
author:
- first_name: Emilie
full_name: Morvant, Emilie
id: 4BAC2A72-F248-11E8-B48F-1D18A9856A87
last_name: Morvant
orcid: 0000-0002-8301-7240
citation:
ama: Morvant E. Domain Adaptation of Weighted Majority Votes via Perturbed Variation-Based
Self-Labeling. Pattern Recognition Letters. 2014;51:37-43. doi:10.1016/j.patrec.2014.08.013
apa: Morvant, E. (2014). Domain Adaptation of Weighted Majority Votes via Perturbed
Variation-Based Self-Labeling. Pattern Recognition Letters. Elsevier. https://doi.org/10.1016/j.patrec.2014.08.013
chicago: Morvant, Emilie. “Domain Adaptation of Weighted Majority Votes via Perturbed
Variation-Based Self-Labeling.” Pattern Recognition Letters. Elsevier,
2014. https://doi.org/10.1016/j.patrec.2014.08.013.
ieee: E. Morvant, “Domain Adaptation of Weighted Majority Votes via Perturbed Variation-Based
Self-Labeling,” Pattern Recognition Letters, vol. 51. Elsevier, pp. 37–43,
2014.
ista: Morvant E. 2014. Domain Adaptation of Weighted Majority Votes via Perturbed
Variation-Based Self-Labeling. Pattern Recognition Letters. 51, 37–43.
mla: Morvant, Emilie. “Domain Adaptation of Weighted Majority Votes via Perturbed
Variation-Based Self-Labeling.” Pattern Recognition Letters, vol. 51, Elsevier,
2014, pp. 37–43, doi:10.1016/j.patrec.2014.08.013.
short: E. Morvant, Pattern Recognition Letters 51 (2014) 37–43.
date_created: 2018-12-11T11:56:05Z
date_published: 2014-10-01T00:00:00Z
date_updated: 2021-01-12T06:55:43Z
day: '01'
doi: 10.1016/j.patrec.2014.08.013
ec_funded: 1
extern: '1'
intvolume: ' 51'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: http://arxiv.org/abs/1410.0334
month: '10'
oa: 1
oa_version: Submitted Version
page: 37-43
project:
- _id: 2532554C-B435-11E9-9278-68D0E5697425
call_identifier: FP7
grant_number: '308036'
name: Lifelong Learning of Visual Scene Understanding
publication: Pattern Recognition Letters
publication_status: published
publisher: Elsevier
publist_id: '4819'
quality_controlled: '1'
status: public
title: Domain Adaptation of Weighted Majority Votes via Perturbed Variation-Based
Self-Labeling
type: journal_article
user_id: 4435EBFC-F248-11E8-B48F-1D18A9856A87
volume: 51
year: '2014'
...
---
_id: '2180'
abstract:
- lang: eng
text: Weighted majority votes allow one to combine the output of several classifiers
or voters. MinCq is a recent algorithm for optimizing the weight of each voter
based on the minimization of a theoretical bound over the risk of the vote with
elegant PAC-Bayesian generalization guarantees. However, while it has demonstrated
good performance when combining weak classifiers, MinCq cannot make use of the
useful a priori knowledge that one may have when using a mixture of weak and strong
voters. In this paper, we propose P-MinCq, an extension of MinCq that can incorporate
such knowledge in the form of a constraint over the distribution of the weights,
along with general proofs of convergence that stand in the sample compression
setting for data-dependent voters. The approach is applied to a vote of k-NN classifiers
with a specific modeling of the voters' performance. P-MinCq significantly outperforms
the classic k-NN classifier, a symmetric NN and MinCq using the same voters. We
show that it is also competitive with LMNN, a popular metric learning algorithm,
and that combining both approaches further reduces the error.
acknowledgement: 'This work was funded by the French project SoLSTiCe ANR-13-BS02-01
of the ANR. '
author:
- first_name: Aurélien
full_name: Bellet, Aurélien
last_name: Bellet
- first_name: Amaury
full_name: Habrard, Amaury
last_name: Habrard
- first_name: Emilie
full_name: Morvant, Emilie
id: 4BAC2A72-F248-11E8-B48F-1D18A9856A87
last_name: Morvant
orcid: 0000-0002-8301-7240
- first_name: Marc
full_name: Sebban, Marc
last_name: Sebban
citation:
ama: Bellet A, Habrard A, Morvant E, Sebban M. Learning a priori constrained weighted
majority votes. Machine Learning. 2014;97(1-2):129-154. doi:10.1007/s10994-014-5462-z
apa: Bellet, A., Habrard, A., Morvant, E., & Sebban, M. (2014). Learning a priori
constrained weighted majority votes. Machine Learning. Springer. https://doi.org/10.1007/s10994-014-5462-z
chicago: Bellet, Aurélien, Amaury Habrard, Emilie Morvant, and Marc Sebban. “Learning
a Priori Constrained Weighted Majority Votes.” Machine Learning. Springer,
2014. https://doi.org/10.1007/s10994-014-5462-z.
ieee: A. Bellet, A. Habrard, E. Morvant, and M. Sebban, “Learning a priori constrained
weighted majority votes,” Machine Learning, vol. 97, no. 1–2. Springer,
pp. 129–154, 2014.
ista: Bellet A, Habrard A, Morvant E, Sebban M. 2014. Learning a priori constrained
weighted majority votes. Machine Learning. 97(1–2), 129–154.
mla: Bellet, Aurélien, et al. “Learning a Priori Constrained Weighted Majority Votes.”
Machine Learning, vol. 97, no. 1–2, Springer, 2014, pp. 129–54, doi:10.1007/s10994-014-5462-z.
short: A. Bellet, A. Habrard, E. Morvant, M. Sebban, Machine Learning 97 (2014)
129–154.
date_created: 2018-12-11T11:56:10Z
date_published: 2014-10-01T00:00:00Z
date_updated: 2021-01-12T06:55:49Z
day: '01'
department:
- _id: ChLa
doi: 10.1007/s10994-014-5462-z
ec_funded: 1
intvolume: ' 97'
issue: 1-2
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://hal.archives-ouvertes.fr/hal-01009578/document
month: '10'
oa: 1
oa_version: Submitted Version
page: 129 - 154
project:
- _id: 2532554C-B435-11E9-9278-68D0E5697425
call_identifier: FP7
grant_number: '308036'
name: Lifelong Learning of Visual Scene Understanding
publication: Machine Learning
publication_status: published
publisher: Springer
publist_id: '4802'
quality_controlled: '1'
scopus_import: 1
status: public
title: Learning a priori constrained weighted majority votes
type: journal_article
user_id: 4435EBFC-F248-11E8-B48F-1D18A9856A87
volume: 97
year: '2014'
...
---
_id: '2189'
abstract:
- lang: fre
text: En apprentissage automatique, nous parlons d'adaptation de domaine lorsque
les données de test (cibles) et d'apprentissage (sources) sont générées selon
différentes distributions. Nous devons donc développer des algorithmes de classification
capables de s'adapter à une nouvelle distribution, pour laquelle aucune information
sur les étiquettes n'est disponible. Nous attaquons cette problématique sous l'angle
de l'approche PAC-Bayésienne qui se focalise sur l'apprentissage de modèles définis
comme des votes de majorité sur un ensemble de fonctions. Dans ce contexte, nous
introduisons PV-MinCq une version adaptative de l'algorithme (non adaptatif) MinCq.
PV-MinCq suit le principe suivant. Nous transférons les étiquettes sources aux
points cibles proches pour ensuite appliquer MinCq sur l'échantillon cible ``auto-étiqueté''
(justifié par une borne théorique). Plus précisément, nous définissons un auto-étiquetage
non itératif qui se focalise dans les régions où les distributions marginales
source et cible sont les plus similaires. Dans un second temps, nous étudions
l'influence de notre auto-étiquetage pour en déduire une procédure de validation
des hyperparamètres. Finalement, notre approche montre des résultats empiriques
prometteurs.
article_processing_charge: No
author:
- first_name: Emilie
full_name: Morvant, Emilie
id: 4BAC2A72-F248-11E8-B48F-1D18A9856A87
last_name: Morvant
orcid: 0000-0002-8301-7240
citation:
ama: 'Morvant E. Adaptation de domaine de vote de majorité par auto-étiquetage non
itératif. In: Vol 1. Elsevier; 2014:49-58.'
apa: 'Morvant, E. (2014). Adaptation de domaine de vote de majorité par auto-étiquetage
non itératif (Vol. 1, pp. 49–58). Presented at the CAP: Conférence Francophone
sur l’Apprentissage Automatique (Machine Learning French Conference), Saint-Etienne,
France: Elsevier.'
chicago: Morvant, Emilie. “Adaptation de Domaine de Vote de Majorité Par Auto-Étiquetage
Non Itératif,” 1:49–58. Elsevier, 2014.
ieee: 'E. Morvant, “Adaptation de domaine de vote de majorité par auto-étiquetage
non itératif,” presented at the CAP: Conférence Francophone sur l’Apprentissage
Automatique (Machine Learning French Conference), Saint-Etienne, France, 2014,
vol. 1, pp. 49–58.'
ista: 'Morvant E. 2014. Adaptation de domaine de vote de majorité par auto-étiquetage
non itératif. CAP: Conférence Francophone sur l’Apprentissage Automatique (Machine
Learning French Conference) vol. 1, 49–58.'
mla: Morvant, Emilie. Adaptation de Domaine de Vote de Majorité Par Auto-Étiquetage
Non Itératif. Vol. 1, Elsevier, 2014, pp. 49–58.
short: E. Morvant, in:, Elsevier, 2014, pp. 49–58.
conference:
location: Saint-Etienne, France
name: 'CAP: Conférence Francophone sur l''Apprentissage Automatique (Machine Learning
French Conference)'
date_created: 2018-12-11T11:56:13Z
date_published: 2014-07-01T00:00:00Z
date_updated: 2021-01-12T06:55:52Z
day: '01'
department:
- _id: ChLa
intvolume: ' 1'
language:
- iso: eng
main_file_link:
- open_access: '1'
url: https://hal.archives-ouvertes.fr/hal-01005776/
month: '07'
oa: 1
oa_version: Preprint
page: 49-58
publication_status: published
publisher: Elsevier
publist_id: '4785'
quality_controlled: '1'
status: public
title: Adaptation de domaine de vote de majorité par auto-étiquetage non itératif
type: conference
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
volume: 1
year: '2014'
...