High-dimensional data can be converted to low-dimensional codes by
training a multilayer neural network with a small central layer to
reconstruct high-dimensional input vectors. Gradient descent can
be used for fine-tuning the weights in such äutoencoder" networks,
but this works well only if the initial weights are close to a good
solution. We describe an effective way of initializing the weights
that allows deep autoencoder networks to learn low-dimensional codes
that work much better than principal components analysis as a tool
to reduce the dimensionality of data.
%0 Journal Article
%1 Hinton:2006
%A Hinton, G. E.
%A Salakhutdinov, R. R.
%D 2006
%J Science
%K imported
%P 504 - 507
%T Reducing the Dimensionality of Data with Neural Networks
%V 313
%X High-dimensional data can be converted to low-dimensional codes by
training a multilayer neural network with a small central layer to
reconstruct high-dimensional input vectors. Gradient descent can
be used for fine-tuning the weights in such äutoencoder" networks,
but this works well only if the initial weights are close to a good
solution. We describe an effective way of initializing the weights
that allows deep autoencoder networks to learn low-dimensional codes
that work much better than principal components analysis as a tool
to reduce the dimensionality of data.
@article{Hinton:2006,
abstract = {High-dimensional data can be converted to low-dimensional codes by
training a multilayer neural network with a small central layer to
reconstruct high-dimensional input vectors. Gradient descent can
be used for fine-tuning the weights in such "autoencoder" networks,
but this works well only if the initial weights are close to a good
solution. We describe an effective way of initializing the weights
that allows deep autoencoder networks to learn low-dimensional codes
that work much better than principal components analysis as a tool
to reduce the dimensionality of data.},
added-at = {2009-06-26T15:25:19.000+0200},
author = {Hinton, G. E. and Salakhutdinov, R. R.},
biburl = {https://www.bibsonomy.org/bibtex/2ec11ecb333fe08f911354cd02e9e4908/butz},
description = {diverse cognitive systems bib},
interhash = {019918b82518b74f443a22dc58a0117f},
intrahash = {ec11ecb333fe08f911354cd02e9e4908},
journal = {Science},
keywords = {imported},
owner = {butz},
pages = {504 - 507},
timestamp = {2009-06-26T15:25:34.000+0200},
title = {Reducing the Dimensionality of Data with Neural Networks},
volume = 313,
year = 2006
}