Note to users. If you're seeing this message, it means that your browser cannot find this page's style/presentation instructions -- or possibly that you are using a browser that does not support current Web standards. Find out more about why this message is appearing, and what you can do to make your experience of our site the best it can be.

Site Tools

  • AAAS
  • Subscribe
  • Feedback

Site Search

Search Advanced

Science 28 July 2006:
Vol. 313. no. 5786, pp. 504 - 507
DOI: 10.1126/science.1127647
Prev | Table of Contents | Next

Reports

Reducing the Dimensionality of Data with Neural Networks

G. E. Hinton* and R. R. Salakhutdinov

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.

Department of Computer Science, University of Toronto, 6 King's College Road, Toronto, Ontario M5S 3G4, Canada.

* To whom correspondence should be addressed; E-mail: hinton{at}cs.toronto.edu

Read the Full Text

To Advertise     Find Products

Science. ISSN 0036-8075 (print), 1095-9203 (online)