%0 Conference Paper %1 VincentPLarochelleH2008 %A Vincent, Pascal %A Larochelle, Hugo %A Bengio, Yoshua %A Manzagol, Pierre-Antoine %B Proceedings of the Twenty-fifth International Conference on Machine Learning (ICML'08) %D 2008 %E Cohen, William W. %E McCallum, Andrew %E Roweis, Sam T. %I ACM %K autoencoder imported ma-zehe neuralnets %P 1096--1103 %T Extracting and Composing Robust Features with Denoising Autoencoders %X Recently, many applications for Restricted Boltzmann Machines (RBMs) have been developed for a large variety of learning problems. However, RBMs are usually used as feature extractors for another learning algorithm or to provide a good initialization for deep feed-forward neural network classifiers, and are not considered as a standalone solution to classification problems. In this paper, we argue that RBMs provide a self-contained framework for deriving competitive non-linear classifiers. We present an evaluation of different learning algorithms for RBMs which aim at introducing a discriminative component to RBM training and improve their performance as classifiers. This approach is simple in that RBMs are used directly to build a classifier, rather than as a stepping stone. Finally, we demonstrate how discriminative RBMs can also be successfully employed in a semi-supervised setting.

@inproceedings{VincentPLarochelleH2008, abstract = {Recently, many applications for Restricted Boltzmann Machines (RBMs) have been developed for a large variety of learning problems. However, RBMs are usually used as feature extractors for another learning algorithm or to provide a good initialization for deep feed-forward neural network classifiers, and are not considered as a standalone solution to classification problems. In this paper, we argue that RBMs provide a self-contained framework for deriving competitive non-linear classifiers. We present an evaluation of different learning algorithms for RBMs which aim at introducing a discriminative component to RBM training and improve their performance as classifiers. This approach is simple in that RBMs are used directly to build a classifier, rather than as a stepping stone. Finally, we demonstrate how discriminative RBMs can also be successfully employed in a semi-supervised setting.}, added-at = {2016-11-19T16:12:22.000+0100}, author = {Vincent, Pascal and Larochelle, Hugo and Bengio, Yoshua and Manzagol, Pierre-Antoine}, biburl = {https://www.bibsonomy.org/bibtex/268b3f10785bab04ebbb21ae21587d2bc/albinzehe}, booktitle = {Proceedings of the Twenty-fifth International Conference on Machine Learning (ICML'08)}, crossref = {ICML08}, editor = {Cohen, William W. and McCallum, Andrew and Roweis, Sam T.}, interhash = {f53f01391a871310794be6721ca56fc8}, intrahash = {68b3f10785bab04ebbb21ae21587d2bc}, keywords = {autoencoder imported ma-zehe neuralnets}, pages = {1096--1103}, publisher = {ACM}, timestamp = {2016-11-19T16:13:11.000+0100}, title = {Extracting and Composing Robust Features with Denoising Autoencoders}, year = 2008 }