Theoretical results suggest that in order to learn the kind of complicated functions that can represent high-level abstractions (e.g., in vision, language, and other AI-level tasks), one may need deep architectures. Deep architectures are composed of multiple levels of non-linear operations, such as in neural nets with many hidden layers or in complicated propositional formulae re-using many sub-formulae. Searching the parameter space of deep architectures is a difficult task, but learning algorithms such as those for Deep Belief Networks have recently been proposed to tackle this problem with notable success, beating the state-of-the-art in certain areas. This monograph discusses the motivations and principles regarding learning algorithms for deep architectures, in particular those exploiting as building blocks unsupervised learning of single-layer models such as Restricted Boltzmann Machines, used to construct deeper models such as Deep Belief Networks.
%0 Journal Article
%1 bengio2009learning
%A Bengio, Y.
%C Hanover, MA, USA
%D 2009
%I Now Publishers Inc.
%J Foundations and Trends® in Machine Learning
%K imported
%N 1
%P 1--127
%R 10.1561/2200000006
%T Learning Deep Architectures for AI
%U http://dx.doi.org/10.1561/2200000006
%V 2
%X Theoretical results suggest that in order to learn the kind of complicated functions that can represent high-level abstractions (e.g., in vision, language, and other AI-level tasks), one may need deep architectures. Deep architectures are composed of multiple levels of non-linear operations, such as in neural nets with many hidden layers or in complicated propositional formulae re-using many sub-formulae. Searching the parameter space of deep architectures is a difficult task, but learning algorithms such as those for Deep Belief Networks have recently been proposed to tackle this problem with notable success, beating the state-of-the-art in certain areas. This monograph discusses the motivations and principles regarding learning algorithms for deep architectures, in particular those exploiting as building blocks unsupervised learning of single-layer models such as Restricted Boltzmann Machines, used to construct deeper models such as Deep Belief Networks.
@article{bengio2009learning,
abstract = {{Theoretical results suggest that in order to learn the kind of complicated functions that can represent high-level abstractions (e.g., in vision, language, and other AI-level tasks), one may need deep architectures. Deep architectures are composed of multiple levels of non-linear operations, such as in neural nets with many hidden layers or in complicated propositional formulae re-using many sub-formulae. Searching the parameter space of deep architectures is a difficult task, but learning algorithms such as those for Deep Belief Networks have recently been proposed to tackle this problem with notable success, beating the state-of-the-art in certain areas. This monograph discusses the motivations and principles regarding learning algorithms for deep architectures, in particular those exploiting as building blocks unsupervised learning of single-layer models such as Restricted Boltzmann Machines, used to construct deeper models such as Deep Belief Networks.}},
added-at = {2017-07-19T15:29:59.000+0200},
address = {Hanover, MA, USA},
author = {Bengio, Y.},
biburl = {https://www.bibsonomy.org/bibtex/2d5e475e8615951bff7e0f521f422f1c1/andreashdez},
citeulike-article-id = {7170021},
citeulike-linkout-0 = {http://portal.acm.org/citation.cfm?id=1658423.1658424},
citeulike-linkout-1 = {http://dx.doi.org/10.1561/2200000006},
doi = {10.1561/2200000006},
interhash = {30174ec5e2667a039cdc30c5d359dc47},
intrahash = {d5e475e8615951bff7e0f521f422f1c1},
issn = {1935-8237},
journal = {Foundations and Trends® in Machine Learning},
keywords = {imported},
month = jan,
number = 1,
pages = {1--127},
posted-at = {2016-04-29 18:48:08},
priority = {0},
publisher = {Now Publishers Inc.},
timestamp = {2017-07-19T15:31:02.000+0200},
title = {{Learning Deep Architectures for AI}},
url = {http://dx.doi.org/10.1561/2200000006},
volume = 2,
year = 2009
}