Stacked convolutional auto-encoders for hierarchical feature extraction

Jonathan Masci, Ueli Meier, Dan Cireşan, Jürgen Schmidhuber

Research output: Chapter in Book/Report/Conference proceedingConference contribution

1567 Scopus citations


We present a novel convolutional auto-encoder (CAE) for unsupervised feature learning. A stack of CAEs forms a convolutional neural network (CNN). Each CAE is trained using conventional on-line gradient descent without additional regularization terms. A max-pooling layer is essential to learn biologically plausible features consistent with those found by previous approaches. Initializing a CNN with filters of a trained CAE stack yields superior performance on a digit (MNIST) and an object recognition (CIFAR10) benchmark. © 2011 Springer-Verlag.
Original languageEnglish (US)
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Number of pages8
StatePublished - Jun 24 2011
Externally publishedYes

Bibliographical note

Generated from Scopus record by KAUST IRTS on 2022-09-14

ASJC Scopus subject areas

  • Theoretical Computer Science
  • General Computer Science


Dive into the research topics of 'Stacked convolutional auto-encoders for hierarchical feature extraction'. Together they form a unique fingerprint.

Cite this