Deep, Big, Simple Neural Nets for Handwritten Digit Recognition

被引：584

作者：

Ciresan, Dan Claudiu ^{[1
]}

Meier, Ueli

Gambardella, Luca Maria

Schmidhuber, Juergen

机构：

[1] IDSIA, CH-6928 Lugano, Switzerland

来源：

NEURAL COMPUTATION | 2010年 / 22卷 / 12期

关键词：

D O I：

10.1162/NECO_a_00052

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

Good old online backpropagation for plain multilayer perceptrons yields a very low 0.35% error rate on the MNIST handwritten digits benchmark. All we need to achieve this best result so far are many hidden layers, many neurons per layer, numerous deformed training images to avoid overfitting, and graphics cards to greatly speed up learning.

引用

页码：3207 / 3220

页数：14

共 24 条

[1]

[Anonymous], 1991, Untersuchungen zu dynamischen neuronalen Netzen[J]

[2]

[Anonymous], 2007, AISTATS

[3]

[Anonymous], P NIPS 2009 WORKSH D

[4]

[Anonymous], 2006, Advances in neural information processing systems

[5]

[Anonymous], 1985, Proceedings of Cognitiva

[6]

Bengio Y., ADV NEURAL INFORM PR, P153

[7]

Chellapilla K., 2006, P ICFHR

[8] Training invariant support vector machines [J].

Decoste, D ;

Schölkopf, B .

MACHINE LEARNING, 2002, 46 (1-3) :161-190

[9] Reducing the dimensionality of data with neural networks [J].

Hinton, G. E. ;

Salakhutdinov, R. R. .

SCIENCE, 2006, 313 (5786) :504-507

[10] To recognize shapes, first learn to generate images [J].

Hinton, Geoffrey E. .

COMPUTATIONAL NEUROSCIENCE: THEORETICAL INSIGHTS INTO BRAIN FUNCTION, 2007, 165 :535-547

← 1 2 3 →