Mercurial > ift6266

comparison writeup/nips2010_submission.tex @ 568:ae6ba0309bf9

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
nouveaux graphes
author Yoshua Bengio <bengioy@iro.umontreal.ca>
date Thu, 03 Jun 2010 13:19:16 -0400
parents b9b811e886ae
children 9d01280ff1c1
comparison
equal deleted inserted replaced
567:08709b62e574 568:ae6ba0309bf9
740 %\subsection{SDA vs MLP vs Humans} 740 %\subsection{SDA vs MLP vs Humans}
741 %\vspace*{-1mm} 741 %\vspace*{-1mm}
742 The models are either trained on NIST (MLP0 and SDA0), 742 The models are either trained on NIST (MLP0 and SDA0),
743 NISTP (MLP1 and SDA1), or P07 (MLP2 and SDA2), and tested 743 NISTP (MLP1 and SDA1), or P07 (MLP2 and SDA2), and tested
744 on either NIST, NISTP or P07, either on the 62-class task 744 on either NIST, NISTP or P07, either on the 62-class task
745 or on the 10-digits task. 745 or on the 10-digits task. Training (including about half
746 for unsupervised pre-training, for DAs) on the larger
747 datasets takes around one day on a GPU-285.
746 Figure~\ref{fig:error-rates-charts} summarizes the results obtained, 748 Figure~\ref{fig:error-rates-charts} summarizes the results obtained,
747 comparing humans, the three MLPs (MLP0, MLP1, MLP2) and the three SDAs (SDA0, SDA1, 749 comparing humans, the three MLPs (MLP0, MLP1, MLP2) and the three SDAs (SDA0, SDA1,
748 SDA2), along with the previous results on the digits NIST special database 750 SDA2), along with the previous results on the digits NIST special database
749 19 test set from the literature, respectively based on ARTMAP neural 751 19 test set from the literature, respectively based on ARTMAP neural
750 networks ~\citep{Granger+al-2007}, fast nearest-neighbor search 752 networks ~\citep{Granger+al-2007}, fast nearest-neighbor search