On 2013 Jun 13, Robert Tibshirani commented:

This paper represents some of the recent work on "Deep Learning", a machine learning approach which models data in an unsupervised way, with multiple hidden layers. This produces factors that are functions of the inputs, which can then be used as input in a supervised problem. It is exciting stuff, and is the centerpiece, for example, of the Google Brain project. I do think that the approach here may be unnecessarily complicated: more recent, simpler proposals can be found eg in

http://ai.stanford.edu/~quocle/faces_full.pdf

http://static.googleusercontent.com/external_content/untrusted_dlcp/research.google.com/en/us/archive/large_deep_networks_nips2012.pdf

On 2013 Jun 13, Robert Tibshirani commented:

This paper represents some of the recent work on "Deep Learning", a machine learning approach which models data in an unsupervised way, with multiple hidden layers. This produces factors that are functions of the inputs, which can then be used as input in a supervised problem. It is exciting stuff, and is the centerpiece, for example, of the Google Brain project. I do think that the approach here may be unnecessarily complicated: more recent, simpler proposals can be found eg in

http://ai.stanford.edu/~quocle/faces_full.pdf

http://static.googleusercontent.com/external_content/untrusted_dlcp/research.google.com/en/us/archive/large_deep_networks_nips2012.pdf