Efficient Temporal Kernels between Feature Sets for Time Series Classification
Romain Tavenard
(1, 2, 3)
,
Simon Malinowski
(4, 5)
,
Laetitia Chapel
(6, 2)
,
Adeline Bailly
(2, 3, 1)
,
Heider Sanchez
(7)
,
Benjamin Bustos
(7)
1
UR2 -
Université de Rennes 2
2 OBELIX - Environment observation with complex imagery
3 LETG - Rennes - Littoral, Environnement, Télédétection, Géomatique
4 UR - Université de Rennes
5 LinkMedia - Creating and exploiting explicit links between multimedia fragments
6 UBS Vannes - Université de Bretagne Sud - Vannes
7 UCHILE - Universidad de Chile = University of Chile [Santiago]
2 OBELIX - Environment observation with complex imagery
3 LETG - Rennes - Littoral, Environnement, Télédétection, Géomatique
4 UR - Université de Rennes
5 LinkMedia - Creating and exploiting explicit links between multimedia fragments
6 UBS Vannes - Université de Bretagne Sud - Vannes
7 UCHILE - Universidad de Chile = University of Chile [Santiago]
Romain Tavenard
- Fonction : Auteur
- PersonId : 5645
- IdHAL : rtavenar
- ORCID : 0000-0002-1439-8465
- IdRef : 154729507
Simon Malinowski
- Fonction : Auteur
- PersonId : 14415
- IdHAL : simon-malinowski
- IdRef : 12983436X
Laetitia Chapel
- Fonction : Auteur
- PersonId : 740638
- IdHAL : laetitia-chapel
- IdRef : 120235943
Adeline Bailly
- Fonction : Auteur
- PersonId : 8984
- IdHAL : adeline-bailly
- IdRef : 227850564
Résumé
In the time-series classification context, the majority of the most accurate core methods are based on the Bag-of-Words framework, in which sets of local features are first extracted from time series. A dictionary of words is then learned and each time series is finally represented by a histogram of word occurrences. This representation induces a loss of information due to the quantization of features into words as all the time series are represented using the same fixed dictionary. In order to overcome this issue, we introduce in this paper a kernel operating directly on sets of features. Then, we extend it to a time-compliant kernel that allows one to take into account the temporal information. We apply this kernel in the time series classification context. Proposed kernel has a quadratic complexity with the size of input feature sets, which is problematic when dealing with long time series. However, we show that kernel approximation techniques can be used to define a good trade-off between accuracy and complexity. We experimentally demonstrate that the proposed kernel can significantly improve the performance of time series classification algorithms based on Bag-of-Words.
Domaines
Méthodes et statistiquesFormat du dépôt | Fichier |
---|---|
Type de dépôt | Communication dans un congrès |
Titre |
en
Efficient Temporal Kernels between Feature Sets for Time Series Classification
|
Résumé |
en
In the time-series classification context, the majority of the most accurate core methods are based on the Bag-of-Words framework, in which sets of local features are first extracted from time series. A dictionary of words is then learned and each time series is finally represented by a histogram of word occurrences. This representation induces a loss of information due to the quantization of features into words as all the time series are represented using the same fixed dictionary. In order to overcome this issue, we introduce in this paper a kernel operating directly on sets of features. Then, we extend it to a time-compliant kernel that allows one to take into account the temporal information. We apply this kernel in the time series classification context. Proposed kernel has a quadratic complexity with the size of input feature sets, which is problematic when dealing with long time series. However, we show that kernel approximation techniques can be used to define a good trade-off between accuracy and complexity. We experimentally demonstrate that the proposed kernel can significantly improve the performance of time series classification algorithms based on Bag-of-Words.
|
Auteur(s) |
Romain Tavenard
1, 2, 3
, Simon Malinowski
4, 5
, Laetitia Chapel
6, 2
, Adeline Bailly
2, 3, 1
, Heider Sanchez
7
, Benjamin Bustos
7
1
UR2 -
Université de Rennes 2
( 406201 )
- Place du recteur Henri Le Moal - CS 24307 - 35043 Rennes cedex
- France
2
OBELIX -
Environment observation with complex imagery
( 255395 )
- Campus de Tohannic, 56017 Vannes Cedex
- France
3
LETG - Rennes -
Littoral, Environnement, Télédétection, Géomatique
( 3177 )
- Maison de la Recherche Place du Recteur Henri Le Moal 35043 RENNES CEDEX
- France
4
UR -
Université de Rennes
( 105160 )
- Campus de Beaulieu, 263 avenue Général Leclerc, CS 74205, 35042 RENNES CEDEX
- France
5
LinkMedia -
Creating and exploiting explicit links between multimedia fragments
( 491438 )
- Campus de Beaulieu 35042 Rennes cedex
- France
6
UBS Vannes -
Université de Bretagne Sud - Vannes
( 178469 )
- Rue Yves Mainguy - 56017 Vannes cedex
- France
7
UCHILE -
Universidad de Chile = University of Chile [Santiago]
( 142796 )
- Av. Libertador Bernardo O'Higgins 1058, Santiago de Chile
- Chili
|
Langue du document |
Anglais
|
Date de production/écriture |
2017
|
Vulgarisation |
Non
|
Actes |
Oui
|
Comité de lecture |
Oui
|
Invité |
Non
|
Audience |
Internationale
|
Date de publication |
2017
|
Titre du congrès |
European Conference on Machine Learning and Principles and Practice of Knowledge Discovery
|
Date début congrès |
2017-09
|
Ville |
Skopje
|
Pays |
Macédoine du Nord
|
Domaine(s) |
|
Projet(s) ANR |
|
Mots-clés |
en
Time series classification
|
Fichier principal
paper.pdf ( 913.9 Ko
)
Télécharger
supplementary_material.pdf ( 398.15 Ko
)
Télécharger
Origine :
Fichiers produits par l'(les) auteur(s)
Origine :
Fichiers produits par l'(les) auteur(s)
Loading...