Mercurial > ift6266

comparison writeup/aistats2011_cameraready.tex @ 634:54e8958e963b

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
bib
author Yoshua Bengio <bengioy@iro.umontreal.ca>
date Sat, 19 Mar 2011 22:57:48 -0400
parents 510220effb14
children 83d53ffe3f25
comparison
equal deleted inserted replaced
631:510220effb14 634:54e8958e963b
206 To achieve these results, we introduce in the next section a sophisticated system 206 To achieve these results, we introduce in the next section a sophisticated system
207 for stochastically transforming character images and then explain the methodology, 207 for stochastically transforming character images and then explain the methodology,
208 which is based on training with or without these transformed images and testing on 208 which is based on training with or without these transformed images and testing on
209 clean ones. 209 clean ones.
210 Code for generating these transformations as well as for the deep learning 210 Code for generating these transformations as well as for the deep learning
211 algorithms are made available at {\tt http://anonymous.url.net}.%{\tt http://hg.assembla.com/ift6266}. 211 algorithms are made available at
212 {\tt http://hg.assembla.com/ift6266}.
212 213
213 %\vspace*{-3mm} 214 %\vspace*{-3mm}
214 %\newpage 215 %\newpage
215 \section{Perturbed and Transformed Character Images} 216 \section{Perturbed and Transformed Character Images}
216 \label{s:perturbations} 217 \label{s:perturbations}
224 Although character transformations have been used before to 225 Although character transformations have been used before to
225 improve character recognizers, this effort is on a large scale both 226 improve character recognizers, this effort is on a large scale both
226 in number of classes and in the complexity of the transformations, hence 227 in number of classes and in the complexity of the transformations, hence
227 in the complexity of the learning task. 228 in the complexity of the learning task.
228 The code for these transformations (mostly Python) is available at 229 The code for these transformations (mostly Python) is available at
229 {\tt http://anonymous.url.net}. All the modules in the pipeline (Figure~\ref{fig:transform}) share 230 {\tt http://hg.assembla.com/ift6266}. All the modules in the pipeline (Figure~\ref{fig:transform}) share
230 a global control parameter ($0 \le complexity \le 1$) that allows one to modulate the 231 a global control parameter ($0 \le complexity \le 1$) that allows one to modulate the
231 amount of deformation or noise introduced. 232 amount of deformation or noise introduced.
232 There are two main parts in the pipeline. The first one, 233 There are two main parts in the pipeline. The first one,
233 from thickness to pinch, performs transformations. The second 234 from thickness to pinch, performs transformations. The second
234 part, from blur to contrast, adds different kinds of noise. 235 part, from blur to contrast, adds different kinds of noise.
235 More details can be found in~\citep{ift6266-tr-anonymous}. 236 More details can be found in~\citep{ARXIV-2010}.
236 237
237 \begin{figure*}[ht] 238 \begin{figure*}[ht]
238 \centering 239 \centering
239 \subfigure[Original]{\includegraphics[scale=0.6]{images/Original.png}\label{fig:torig}} 240 \subfigure[Original]{\includegraphics[scale=0.6]{images/Original.png}\label{fig:torig}}
240 \subfigure[Thickness]{\includegraphics[scale=0.6]{images/Thick_only.png}} 241 \subfigure[Thickness]{\includegraphics[scale=0.6]{images/Thick_only.png}}
799 does not allow the shallow or purely supervised models to discover 800 does not allow the shallow or purely supervised models to discover
800 the kind of better basins associated 801 the kind of better basins associated
801 with deep learning and out-of-distribution examples. 802 with deep learning and out-of-distribution examples.
802 803
803 A Flash demo of the recognizer (where both the MLP and the SDA can be compared) 804 A Flash demo of the recognizer (where both the MLP and the SDA can be compared)
804 can be executed on-line at the anonymous site {\tt http://deep.host22.com}. 805 can be executed on-line at {\tt http://deep.host22.com}.
805 806
806 \iffalse 807 \iffalse
807 \section*{Appendix I: Detailed Numerical Results} 808 \section*{Appendix I: Detailed Numerical Results}
808 809
809 These tables correspond to Figures 2 and 3 and contain the raw error rates for each model and dataset considered. 810 These tables correspond to Figures 2 and 3 and contain the raw error rates for each model and dataset considered.