Committees of deep feedforward networks trained with few data

Bogdan Miclut; Thomas Kaester; Thomas Martinetz; Erhardt Barth

Committees of deep feedforward networks trained with few data

Bogdan Miclut, Thomas Kaester, Thomas Martinetz, Erhardt Barth

Institut für Neuro- und Bioinformatik

Abstract

Deep convolutional neural networks are known to give good results on image classification tasks. In this paper we present a method to improve the classification result by combining multiple such networks in a committee. We adopt the STL-10 dataset which has very few training examples and show that our method can achieve results that are better than the state of the art. The networks are trained layer-wise and no backpropagation is used. We also explore the effects of dataset augmentation by mirroring, rotation, and scaling.

Originalsprache	Englisch
Zeitschrift	arXiv.org
Seitenumfang	7
Publikationsstatus	Veröffentlicht - 23.06.2014

UN SDGs

Dieser Output leistet einen Beitrag zu folgendem(n) Ziel(en) für nachhaltige Entwicklung

Dokumente

1406.5947v1Endgültige, publizierte Fassung (Verlagsversion), 112 KB

https://arxiv.org/abs/1406.5947

Zitieren

@article{f36ea0e3dce84ed08617fb6a2184df6e,

title = "Committees of deep feedforward networks trained with few data",

abstract = "Deep convolutional neural networks are known to give good results on image classification tasks. In this paper we present a method to improve the classification result by combining multiple such networks in a committee. We adopt the STL-10 dataset which has very few training examples and show that our method can achieve results that are better than the state of the art. The networks are trained layer-wise and no backpropagation is used. We also explore the effects of dataset augmentation by mirroring, rotation, and scaling.",

author = "Bogdan Miclut and Thomas Kaester and Thomas Martinetz and Erhardt Barth",

year = "2014",

month = jun,

day = "23",

language = "English",

journal = "arXiv.org",

}

TY - JOUR

T1 - Committees of deep feedforward networks trained with few data

AU - Miclut, Bogdan

AU - Kaester, Thomas

AU - Martinetz, Thomas

AU - Barth, Erhardt

PY - 2014/6/23

Y1 - 2014/6/23

N2 - Deep convolutional neural networks are known to give good results on image classification tasks. In this paper we present a method to improve the classification result by combining multiple such networks in a committee. We adopt the STL-10 dataset which has very few training examples and show that our method can achieve results that are better than the state of the art. The networks are trained layer-wise and no backpropagation is used. We also explore the effects of dataset augmentation by mirroring, rotation, and scaling.

AB - Deep convolutional neural networks are known to give good results on image classification tasks. In this paper we present a method to improve the classification result by combining multiple such networks in a committee. We adopt the STL-10 dataset which has very few training examples and show that our method can achieve results that are better than the state of the art. The networks are trained layer-wise and no backpropagation is used. We also explore the effects of dataset augmentation by mirroring, rotation, and scaling.

M3 - Journal articles

JO - arXiv.org

JF - arXiv.org

ER -

Committees of deep feedforward networks trained with few data

Abstract

UN SDGs

Dokumente

Fingerprint

Zitieren