Real adaboost with gate controlled fusion

Efrain Tito Mayhua Lopez; Vanessa Gomez-Verdejo; Aníbal R. Figueiras-Vidal

doi:10.1109/TNNLS.2012.2219318

Real adaboost with gate controlled fusion

Efrain Tito Mayhua Lopez, Vanessa Gomez-Verdejo, Aníbal R. Figueiras-Vidal

Resultado de la investigación: Contribución a una revista › Artículo › revisión exhaustiva

10 Citas (Scopus)

Resumen

In this brief, we propose to increase the capabilities of standard real AdaBoost (RAB) architectures by replacing their linear combinations with a fusion controlled by a gate with fixed kernels. Experimental results in a series of well-known benchmark problems support the effectiveness of this approach in improving classification performance. Although the need for cross-validation processes obviously leads to higher training requirements and more computational effort, the operation load is never much higher; in many cases it is even lower than that of competitive RAB schemes.

Idioma original	Inglés
Número de artículo	6352922
Páginas (desde-hasta)	2003-2009
Número de páginas	7
Publicación	IEEE Transactions on Neural Networks and Learning Systems
Volumen	23
N.º	12
DOI	https://doi.org/10.1109/TNNLS.2012.2219318
Estado	Publicada - 2012
Publicado de forma externa	Sí

Nota bibliográfica

Funding Information:
Manuscript received July 5, 2011; revised September 10, 2012; accepted September 10, 2012. Date of publication November 10, 2012; date of current version November 20, 2012. This work was supported in part by the Spanish MICINN under Grant TEC 2011-22480, Grant TIN 2011-24533, and Grant PRI-PIBIN 2011-1266.

Acceder al documento

10.1109/TNNLS.2012.2219318

Otros archivos y enlaces

Enlace a la publicación en Scopus

Citar esto

@article{3066c76d2d73474693005a113080de8c,

title = "Real adaboost with gate controlled fusion",

abstract = "In this brief, we propose to increase the capabilities of standard real AdaBoost (RAB) architectures by replacing their linear combinations with a fusion controlled by a gate with fixed kernels. Experimental results in a series of well-known benchmark problems support the effectiveness of this approach in improving classification performance. Although the need for cross-validation processes obviously leads to higher training requirements and more computational effort, the operation load is never much higher; in many cases it is even lower than that of competitive RAB schemes.",

keywords = "Classification, ensembles, mixtures of experts, neural networks, real AdaBoost",

author = "{Mayhua Lopez}, {Efrain Tito} and Vanessa Gomez-Verdejo and Figueiras-Vidal, {An{\'i}bal R.}",

note = "Funding Information: Manuscript received July 5, 2011; revised September 10, 2012; accepted September 10, 2012. Date of publication November 10, 2012; date of current version November 20, 2012. This work was supported in part by the Spanish MICINN under Grant TEC 2011-22480, Grant TIN 2011-24533, and Grant PRI-PIBIN 2011-1266.",

year = "2012",

doi = "10.1109/TNNLS.2012.2219318",

language = "Ingl{\'e}s",

volume = "23",

pages = "2003--2009",

journal = "IEEE Transactions on Neural Networks",

issn = "2162-237X",

publisher = "IEEE Computational Intelligence Society",

number = "12",

}

TY - JOUR

T1 - Real adaboost with gate controlled fusion

AU - Mayhua Lopez, Efrain Tito

AU - Gomez-Verdejo, Vanessa

AU - Figueiras-Vidal, Aníbal R.

N1 - Funding Information: Manuscript received July 5, 2011; revised September 10, 2012; accepted September 10, 2012. Date of publication November 10, 2012; date of current version November 20, 2012. This work was supported in part by the Spanish MICINN under Grant TEC 2011-22480, Grant TIN 2011-24533, and Grant PRI-PIBIN 2011-1266.

PY - 2012

Y1 - 2012

N2 - In this brief, we propose to increase the capabilities of standard real AdaBoost (RAB) architectures by replacing their linear combinations with a fusion controlled by a gate with fixed kernels. Experimental results in a series of well-known benchmark problems support the effectiveness of this approach in improving classification performance. Although the need for cross-validation processes obviously leads to higher training requirements and more computational effort, the operation load is never much higher; in many cases it is even lower than that of competitive RAB schemes.

AB - In this brief, we propose to increase the capabilities of standard real AdaBoost (RAB) architectures by replacing their linear combinations with a fusion controlled by a gate with fixed kernels. Experimental results in a series of well-known benchmark problems support the effectiveness of this approach in improving classification performance. Although the need for cross-validation processes obviously leads to higher training requirements and more computational effort, the operation load is never much higher; in many cases it is even lower than that of competitive RAB schemes.

KW - Classification

KW - ensembles

KW - mixtures of experts

KW - neural networks

KW - real AdaBoost

UR - http://www.scopus.com/inward/record.url?scp=84876902833&partnerID=8YFLogxK

U2 - 10.1109/TNNLS.2012.2219318

DO - 10.1109/TNNLS.2012.2219318

M3 - Artículo

AN - SCOPUS:84876902833

VL - 23

SP - 2003

EP - 2009

JO - IEEE Transactions on Neural Networks

JF - IEEE Transactions on Neural Networks

SN - 2162-237X

IS - 12

M1 - 6352922

ER -

Real adaboost with gate controlled fusion

Resumen

Nota bibliográfica

Acceder al documento

Otros archivos y enlaces

Huella

Citar esto