annotate writeup/techreport.tex @ 422:e7790db265b1

Basic text for section 3, add a bit more detail to section 4.2.2
author Arnaud Bergeron <abergeron@gmail.com>
date Fri, 30 Apr 2010 16:24:30 -0400
parents 0282882aa91f
children e4eb3ee7a0cf
rev   line source
379
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
1 \documentclass[12pt,letterpaper]{article}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
2 \usepackage[utf8]{inputenc}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
3 \usepackage{graphicx}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
4 \usepackage{times}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
5 \usepackage{mlapa}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
6
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
7 \begin{document}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
8 \title{Generating and Exploiting Perturbed Training Data for Deep Architectures}
381
0a91fc69ff90 authors
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 379
diff changeset
9 \author{The IFT6266 Gang}
379
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
10 \date{April 2010, Technical Report, Dept. IRO, U. Montreal}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
11
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
12 \maketitle
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
13
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
14 \begin{abstract}
392
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
15 Recent theoretical and empirical work in statistical machine learning has
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
16 demonstrated the importance of learning algorithms for deep
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
17 architectures, i.e., function classes obtained by composing multiple
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
18 non-linear transformations. In the area of handwriting recognition,
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
19 deep learning algorithms
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
20 had been evaluated on rather small datasets with a few tens of thousands
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
21 of examples. Here we propose a powerful generator of variations
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
22 of examples for character images based on a pipeline of stochastic
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
23 transformations that include not only the usual affine transformations
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
24 but also the addition of slant, local elastic deformations, changes
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
25 in thickness, background images, color, contrast, occlusion, and
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
26 various types of pixel and spatially correlated noise.
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
27 We evaluate a deep learning algorithm (Stacked Denoising Autoencoders)
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
28 on the task of learning to classify digits and letters transformed
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
29 with this pipeline, using the hundreds of millions of generated examples
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
30 and testing on the full NIST test set.
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
31 We find that the SDA outperforms its
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
32 shallow counterpart, an ordinary Multi-Layer Perceptron,
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
33 and that it is better able to take advantage of the additional
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
34 generated data.
379
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
35 \end{abstract}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
36
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
37 \section{Introduction}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
38
392
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
39 Deep Learning has emerged as a promising new area of research in
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
40 statistical machine learning (see~\emcite{Bengio-2009} for a review).
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
41 Learning algorithms for deep architectures are centered on the learning
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
42 of useful representations of data, which are better suited to the task at hand.
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
43 This is in great part inspired by observations of the mammalian visual cortex,
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
44 which consists of a chain of processing elements, each of which is associated with a
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
45 different representation. In fact,
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
46 it was found recently that the features learnt in deep architectures resemble
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
47 those observed in the first two of these stages (in areas V1 and V2
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
48 of visual cortex)~\cite{HonglakL2008}.
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
49 Processing images typically involves transforming the raw pixel data into
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
50 new {\bf representations} that can be used for analysis or classification.
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
51 For example, a principal component analysis representation linearly projects
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
52 the input image into a lower-dimensional feature space.
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
53 Why learn a representation? Current practice in the computer vision
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
54 literature converts the raw pixels into a hand-crafted representation
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
55 (e.g.\ SIFT features~\cite{Lowe04}), but deep learning algorithms
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
56 tend to discover similar features in their first few
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
57 levels~\cite{HonglakL2008,ranzato-08,Koray-08,VincentPLarochelleH2008-very-small}.
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
58 Learning increases the
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
59 ease and practicality of developing representations that are at once
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
60 tailored to specific tasks, yet are able to borrow statistical strength
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
61 from other related tasks (e.g., modeling different kinds of objects). Finally, learning the
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
62 feature representation can lead to higher-level (more abstract, more
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
63 general) features that are more robust to unanticipated sources of
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
64 variance extant in real data.
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
65
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
66 Whereas a deep architecture can in principle be more powerful than a shallow
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
67 one in terms of representation, depth appears to render the training problem
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
68 more difficult in terms of optimization and local minima.
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
69 It is also only recently that
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
70 successful algorithms were proposed to overcome some of these
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
71 difficulties.
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
72
379
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
73 \section{Perturbation and Transformation of Character Images}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
74
407
fe2e2964e7a3 description des transformations en cours ajout d un fichier special.bib pour des references specifiques
Xavier Glorot <glorotxa@iro.umontreal.ca>
parents: 393
diff changeset
75 \subsection{Adding Slant}
fe2e2964e7a3 description des transformations en cours ajout d un fichier special.bib pour des references specifiques
Xavier Glorot <glorotxa@iro.umontreal.ca>
parents: 393
diff changeset
76 In order to mimic a slant effect, we simply shift each row of the image proportionnaly to its height.
fe2e2964e7a3 description des transformations en cours ajout d un fichier special.bib pour des references specifiques
Xavier Glorot <glorotxa@iro.umontreal.ca>
parents: 393
diff changeset
77 The coefficient is randomly sampled according to the complexity level and can be negatif or positif with equal probability.
fe2e2964e7a3 description des transformations en cours ajout d un fichier special.bib pour des references specifiques
Xavier Glorot <glorotxa@iro.umontreal.ca>
parents: 393
diff changeset
78
fe2e2964e7a3 description des transformations en cours ajout d un fichier special.bib pour des references specifiques
Xavier Glorot <glorotxa@iro.umontreal.ca>
parents: 393
diff changeset
79 \subsection{Changing Thickness}
fe2e2964e7a3 description des transformations en cours ajout d un fichier special.bib pour des references specifiques
Xavier Glorot <glorotxa@iro.umontreal.ca>
parents: 393
diff changeset
80 To change the thickness of the characters we used morpholigical operators: dilation and erosion~\cite{Haralick87,Serra82}.
fe2e2964e7a3 description des transformations en cours ajout d un fichier special.bib pour des references specifiques
Xavier Glorot <glorotxa@iro.umontreal.ca>
parents: 393
diff changeset
81 The basic idea of such transform is, for each pixel, to multiply in the element-wise manner its neighbourhood with a matrix called the structuring element.
fe2e2964e7a3 description des transformations en cours ajout d un fichier special.bib pour des references specifiques
Xavier Glorot <glorotxa@iro.umontreal.ca>
parents: 393
diff changeset
82 Then for dilation we remplace the pixel value by the maximum of the result, or the minimum for erosion.
fe2e2964e7a3 description des transformations en cours ajout d un fichier special.bib pour des references specifiques
Xavier Glorot <glorotxa@iro.umontreal.ca>
parents: 393
diff changeset
83 This will dilate or erode objects in the image, the strength of the transform only depends on the structuring element.
fe2e2964e7a3 description des transformations en cours ajout d un fichier special.bib pour des references specifiques
Xavier Glorot <glorotxa@iro.umontreal.ca>
parents: 393
diff changeset
84 We used ten different structural elements with various shapes (the biggest is $5\times5$).
fe2e2964e7a3 description des transformations en cours ajout d un fichier special.bib pour des references specifiques
Xavier Glorot <glorotxa@iro.umontreal.ca>
parents: 393
diff changeset
85 for each image, we radomly sample the operator type (dilation or erosion) and one structural element
fe2e2964e7a3 description des transformations en cours ajout d un fichier special.bib pour des references specifiques
Xavier Glorot <glorotxa@iro.umontreal.ca>
parents: 393
diff changeset
86 from a subset depending of the complexity (the higher the complexity, the biggest the structural element can be).
fe2e2964e7a3 description des transformations en cours ajout d un fichier special.bib pour des references specifiques
Xavier Glorot <glorotxa@iro.umontreal.ca>
parents: 393
diff changeset
87 Erosion allows only the five smallest structural elements because when the character is too thin it may erase it completly.
fe2e2964e7a3 description des transformations en cours ajout d un fichier special.bib pour des references specifiques
Xavier Glorot <glorotxa@iro.umontreal.ca>
parents: 393
diff changeset
88
379
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
89 \subsection{Affine Transformations}
407
fe2e2964e7a3 description des transformations en cours ajout d un fichier special.bib pour des references specifiques
Xavier Glorot <glorotxa@iro.umontreal.ca>
parents: 393
diff changeset
90 We generate an affine transform matrix according to the complexity level, then we apply it directly to the image.
fe2e2964e7a3 description des transformations en cours ajout d un fichier special.bib pour des references specifiques
Xavier Glorot <glorotxa@iro.umontreal.ca>
parents: 393
diff changeset
91 This allows to produce scaling, translation, rotation and shearing variances. We took care that the maximum rotation applied
fe2e2964e7a3 description des transformations en cours ajout d un fichier special.bib pour des references specifiques
Xavier Glorot <glorotxa@iro.umontreal.ca>
parents: 393
diff changeset
92 to the image is low enough not to confuse classes.
fe2e2964e7a3 description des transformations en cours ajout d un fichier special.bib pour des references specifiques
Xavier Glorot <glorotxa@iro.umontreal.ca>
parents: 393
diff changeset
93
379
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
94 \subsection{Local Elastic Deformations}
415
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
95
416
5f9d04dda707 Correction d'une erreur pour pinch et ajout d'une ref bibliographique
fsavard
parents: 415
diff changeset
96 This filter induces a "wiggly" effect in the image. The description here will be brief, as the algorithm follows precisely what is described in \cite{SimardSP03}.
415
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
97
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
98 The general idea is to generate two "displacements" fields, for horizontal and vertical displacements of pixels. Each of these fields has the same size as the original image.
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
99
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
100 When generating the transformed image, we'll loop over the x and y positions in the fields and select, as a value, the value of the pixel in the original image at the (relative) position given by the displacement fields for this x and y. If the position we'd retrieve is outside the borders of the image, we use a 0 value instead.
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
101
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
102 To generate a pixel in either field, first a value between -1 and 1 is chosen from a uniform distribution. Then all the pixels, in both fields, is multiplied by a constant $\alpha$ which controls the intensity of the displacements (bigger $\alpha$ translates into larger wiggles).
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
103
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
104 As a final step, each field is convoluted with a Gaussian 2D kernel of standard deviation $\sigma$. Visually, this results in a "blur" filter. This has the effect of making values next to each other in the displacement fields similar. In effect, this makes the wiggles more coherent, less noisy.
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
105
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
106 As displacement fields were long to compute, 50 pairs of fields were generated per complexity in increments of 0.1 (50 pairs for 0.1, 50 pairs for 0.2, etc.), and afterwards, given a complexity, we selected randomly among the 50 corresponding pairs.
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
107
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
108 $\sigma$ and $\alpha$ were linked to complexity through the formulas $\alpha = \sqrt[3]{complexity} \times 10.0$ and $\sigma = 10 - 7 \times \sqrt[3]{complexity}$.
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
109
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
110 \subsection{Motion Blur}
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
111
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
112 This is a GIMP filter we applied, a "linear motion blur" in GIMP terminology. The description will be brief as it is a well-known filter.
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
113
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
114 This algorithm has two input parameters, $length$ and $angle$. The value of a pixel in the final image is the mean value of the $length$ first pixels found by moving in the $angle$ direction. An approximation of this idea is used, as we won't fall onto precise pixels by following that direction. This is done using the Bresenham line algorithm.
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
115
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
116 The angle, in our case, is chosen from a uniform distribution over $[0,360]$ degrees. The length, though, depends on the complexity; it's sampled from a Gaussian distribution of mean 0 and standard deviation $\sigma = 3 \times complexity$.
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
117
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
118 \subsection{Pinch}
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
119
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
120 This is another GIMP filter we used. The filter is in fact named "Whirl and pinch", but we don't use the "whirl" part (whirl is set to 0). As described in GIMP, a pinch is "similar to projecting the image onto an elastic surface and pressing or pulling on the center of the surface".
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
121
416
5f9d04dda707 Correction d'une erreur pour pinch et ajout d'une ref bibliographique
fsavard
parents: 415
diff changeset
122 Mathematically, for a square input image, think of drawing a circle of radius $r$ around a center point $C$. Any point (pixel) $P$ belonging to that disk (region inside circle) will have its value recalculated by taking the value of another "source" pixel in the original image. The position of that source pixel is found on the line thats goes through $C$ and $P$, but at some other distance $d_2$. Define $d_1$ to be the distance between $P$ and $C$. $d_2$ is given by $d_2 = sin(\frac{\pi{}d_1}{2r})^{-pinch} \times d_1$, where $pinch$ is a parameter to the filter.
5f9d04dda707 Correction d'une erreur pour pinch et ajout d'une ref bibliographique
fsavard
parents: 415
diff changeset
123
417
0282882aa91f Completed the pinch transformation text
fsavard
parents: 416
diff changeset
124 If the region considered is not square then, before computing $d_2$, the smallest dimension (x or y) is stretched such that we may consider the region as if it was square. Then, after $d_2$ has been computed and corresponding components $d_2\_x$ and $d_2\_y$ have been found, the component corresponding to the stretched dimension is compressed back by an inverse ratio.
415
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
125
417
0282882aa91f Completed the pinch transformation text
fsavard
parents: 416
diff changeset
126 The actual value is given by bilinear interpolation considering the pixels around the (non-integer) source position thus found.
415
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
127
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
128 The value for $pinch$ in our case was given by sampling from an uniform distribution over the range $[-complexity, 0.7 \times complexity]$.
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
129
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
130
379
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
131 \subsection{Occlusion}
415
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
132
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
133 This filter selects random parts of other (hereafter "occlusive") letter images and places them over the original letter (hereafter "occluded") image. To be more precise, having selected a subregion of the occlusive image and a desination position in the occluded image, to determine the final value for a given overlapping pixel, it selects whichever pixel is the lightest. As a reminder, the background value is 0, black, so the value nearest to 1 is selected.
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
134
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
135 To select a subpart of the occlusive image, four numbers are generated. For compability with the code, we'll call them "haut", "bas", "gauche" and "droite" (respectively meaning top, bottom, left and right). Each of these numbers is selected according to a Gaussian distribution of mean $8 \times complexity$ and standard deviation $2$. This means the largest the complexity is, the biggest the occlusion will be. The absolute value is taken, as the numbers must be positive, and the maximum value is capped at 15.
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
136
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
137 These four sizes collectively define a window centered on the middle pixel of the occlusive image. This is the part that will be extracted as the occlusion.
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
138
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
139 The next step is to select a destination position in the occluded image. Vertical and horizontal displacements $y\_arrivee$ and $x\_arrivee$ are selected according to Gaussian distributions of mean 0 and of standard deviations of, respectively, 3 and 2. Then an horizontal placement mode, $endroit$ (meaning location), is selected to be of three values meaning left, middle or right.
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
140
416
5f9d04dda707 Correction d'une erreur pour pinch et ajout d'une ref bibliographique
fsavard
parents: 415
diff changeset
141 If $endroit$ is "middle", the occlusion will be horizontally centered around the horizontal middle of the occluded image, then shifted according to $x\_arrivee$. If $endroit$ is "left", it will be placed on the left of the occluded image, then displaced right according to $x\_arrivee$. The contrary happens if $endroit$ is $right$.
415
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
142
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
143 In both the horizontal and vertical positionning, the maximum position in either direction is such that the selected occlusion won't go beyond the borders of the occluded image.
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
144
416
5f9d04dda707 Correction d'une erreur pour pinch et ajout d'une ref bibliographique
fsavard
parents: 415
diff changeset
145 This filter has a probability of not being applied, at all, of 60\%.
415
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
146
379
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
147 \subsection{Background Images}
415
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
148
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
149 This transformation adds a random background behind the letter. The background is chosen by first selecting, at random, an image from a set of images. Then we choose a 32x32 subregion of that image as the background image (by sampling x and y positions uniformly while making sure not to cross image borders).
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
150
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
151 To combine the original letter image and the background image, contrast adjustments are made. We first get the maximal values (i.e. maximal intensity) for both the original image and the background image, $maximage$ and $maxbg$. We also have a parameter, $contrast$, given by sampling from a uniform distribution over $[complexity, 1]$.
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
152
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
153 Once we have all these numbers, we first adjust the values for the background image. Each pixel value is multiplied by $\frac{max(maximage - contrast, 0)}{maxbg}$. Therefore the higher the contrast, the darkest the background will be.
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
154
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
155 The final image is found by taking the brightest (i.e. value nearest to 1) pixel from either the background image or the corresponding pixel in the original image.
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
156
379
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
157 \subsection{Salt and Pepper Noise}
415
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
158
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
159 This filter adds noise to the image by randomly selecting a certain number of them and, for those selected pixels, assign a random value according to a uniform distribution over the $[0,1]$ ranges. This last distribution does not change according to complexity. Instead, the number of selected pixels does: the proportion of changed pixels corresponds to $complexity / 5$, which means, as a maximum, 20\% of the pixels will be randomized. On the lowest extreme, no pixel is changed.
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
160
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
161 This filter also has a probability of not being applied, at all, of 25\%.
1e9788ce1680 Added the parts concerning the transformations I'd announced I'd do: Local elastic deformations; occlusions; gimp transformations; salt and pepper noise; background images
fsavard
parents: 411
diff changeset
162
379
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
163 \subsection{Spatially Gaussian Noise}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
164 \subsection{Color and Contrast Changes}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
165
393
4c840798d290 added examples of figure and table of results
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 392
diff changeset
166 \begin{figure}[h]
4c840798d290 added examples of figure and table of results
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 392
diff changeset
167 \resizebox{.99\textwidth}{!}{\includegraphics{images/example_t.png}}\\
4c840798d290 added examples of figure and table of results
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 392
diff changeset
168 \caption{Illustration of the pipeline of stochastic
4c840798d290 added examples of figure and table of results
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 392
diff changeset
169 transformations applied to the image of a lower-case t
4c840798d290 added examples of figure and table of results
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 392
diff changeset
170 (the upper left image). Each image in the pipeline (going from
4c840798d290 added examples of figure and table of results
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 392
diff changeset
171 left to right, first top line, then bottom line) shows the result
4c840798d290 added examples of figure and table of results
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 392
diff changeset
172 of applying one of the modules in the pipeline. The last image
4c840798d290 added examples of figure and table of results
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 392
diff changeset
173 (bottom right) is used as training example.}
4c840798d290 added examples of figure and table of results
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 392
diff changeset
174 \label{fig:pipeline}
4c840798d290 added examples of figure and table of results
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 392
diff changeset
175 \end{figure}
379
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
176
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
177 \section{Learning Algorithms for Deep Architectures}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
178
422
e7790db265b1 Basic text for section 3, add a bit more detail to section 4.2.2
Arnaud Bergeron <abergeron@gmail.com>
parents: 417
diff changeset
179 Learning for deep network has long been a problem since well-known learning algorithms do not generalize well on deep architectures.
e7790db265b1 Basic text for section 3, add a bit more detail to section 4.2.2
Arnaud Bergeron <abergeron@gmail.com>
parents: 417
diff changeset
180 Using these training algorithms on deep network usually yields to a worse generalization than on shallow networks.
e7790db265b1 Basic text for section 3, add a bit more detail to section 4.2.2
Arnaud Bergeron <abergeron@gmail.com>
parents: 417
diff changeset
181 Recently, new initialization techniques have been discovered that enable better generalization overall.
e7790db265b1 Basic text for section 3, add a bit more detail to section 4.2.2
Arnaud Bergeron <abergeron@gmail.com>
parents: 417
diff changeset
182
e7790db265b1 Basic text for section 3, add a bit more detail to section 4.2.2
Arnaud Bergeron <abergeron@gmail.com>
parents: 417
diff changeset
183 One of these initialization techniques is denoising auto-encoders.
e7790db265b1 Basic text for section 3, add a bit more detail to section 4.2.2
Arnaud Bergeron <abergeron@gmail.com>
parents: 417
diff changeset
184 The principle is that each layer starting from the bottom is trained to encode and decode their input and the encoding part is kept as initialization for the weights and bias of the network.
e7790db265b1 Basic text for section 3, add a bit more detail to section 4.2.2
Arnaud Bergeron <abergeron@gmail.com>
parents: 417
diff changeset
185 For more details see section \ref{SdA}.
e7790db265b1 Basic text for section 3, add a bit more detail to section 4.2.2
Arnaud Bergeron <abergeron@gmail.com>
parents: 417
diff changeset
186
e7790db265b1 Basic text for section 3, add a bit more detail to section 4.2.2
Arnaud Bergeron <abergeron@gmail.com>
parents: 417
diff changeset
187 After initialization is done, standard training algorithms work.
e7790db265b1 Basic text for section 3, add a bit more detail to section 4.2.2
Arnaud Bergeron <abergeron@gmail.com>
parents: 417
diff changeset
188 In this case, since we have large data sets we use stochastic gradient descent.
e7790db265b1 Basic text for section 3, add a bit more detail to section 4.2.2
Arnaud Bergeron <abergeron@gmail.com>
parents: 417
diff changeset
189 This resemble minibatch training except that the batches are selected at random.
e7790db265b1 Basic text for section 3, add a bit more detail to section 4.2.2
Arnaud Bergeron <abergeron@gmail.com>
parents: 417
diff changeset
190 To speed up computation, we randomly pre-arranged examples in batches and used those for all training experiments.
e7790db265b1 Basic text for section 3, add a bit more detail to section 4.2.2
Arnaud Bergeron <abergeron@gmail.com>
parents: 417
diff changeset
191
379
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
192 \section{Experimental Setup}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
193
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
194 \subsection{Training Datasets}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
195
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
196 \subsubsection{Data Sources}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
197
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
198 \begin{itemize}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
199 \item {\bf NIST}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
200 \item {\bf Fonts}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
201 \item {\bf Captchas}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
202 \item {\bf OCR data}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
203 \end{itemize}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
204
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
205 \subsubsection{Data Sets}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
206 \begin{itemize}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
207 \item {\bf NIST}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
208 \item {\bf P07}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
209 \item {\bf NISTP} {\em ne pas utiliser PNIST mais NISTP, pour rester politically correct...}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
210 \end{itemize}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
211
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
212 \subsection{Models and their Hyperparameters}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
213
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
214 \subsubsection{Multi-Layer Perceptrons (MLP)}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
215
410
6330298791fb Description brève de MLP et SdA
Arnaud Bergeron <abergeron@gmail.com>
parents: 407
diff changeset
216 An MLP is a family of functions that are described by stacking layers of of a function similar to
6330298791fb Description brève de MLP et SdA
Arnaud Bergeron <abergeron@gmail.com>
parents: 407
diff changeset
217 $$g(x) = \tanh(b+Wx)$$
6330298791fb Description brève de MLP et SdA
Arnaud Bergeron <abergeron@gmail.com>
parents: 407
diff changeset
218 The input, $x$, is a $d$-dimension vector.
6330298791fb Description brève de MLP et SdA
Arnaud Bergeron <abergeron@gmail.com>
parents: 407
diff changeset
219 The output, $g(x)$, is a $m$-dimension vector.
411
4f69d915d142 Better description of the model parameters.
Arnaud Bergeron <abergeron@gmail.com>
parents: 410
diff changeset
220 The parameter $W$ is a $m\times d$ matrix and is called the weight matrix.
4f69d915d142 Better description of the model parameters.
Arnaud Bergeron <abergeron@gmail.com>
parents: 410
diff changeset
221 The parameter $b$ is a $m$-vector and is called the bias vector.
410
6330298791fb Description brève de MLP et SdA
Arnaud Bergeron <abergeron@gmail.com>
parents: 407
diff changeset
222 The non-linearity (here $\tanh$) is applied element-wise to the output vector.
6330298791fb Description brève de MLP et SdA
Arnaud Bergeron <abergeron@gmail.com>
parents: 407
diff changeset
223 Usually the input is referred to a input layer and similarly for the output.
6330298791fb Description brève de MLP et SdA
Arnaud Bergeron <abergeron@gmail.com>
parents: 407
diff changeset
224 You can of course chain several such functions to obtain a more complex one.
6330298791fb Description brève de MLP et SdA
Arnaud Bergeron <abergeron@gmail.com>
parents: 407
diff changeset
225 Here is a common example
6330298791fb Description brève de MLP et SdA
Arnaud Bergeron <abergeron@gmail.com>
parents: 407
diff changeset
226 $$f(x) = c + V\tanh(b+Wx)$$
6330298791fb Description brève de MLP et SdA
Arnaud Bergeron <abergeron@gmail.com>
parents: 407
diff changeset
227 In this case the intermediate layer corresponding to $\tanh(b+Wx)$ is called a hidden layer.
6330298791fb Description brève de MLP et SdA
Arnaud Bergeron <abergeron@gmail.com>
parents: 407
diff changeset
228 Here the output layer does not have the same non-linearity as the hidden layer.
6330298791fb Description brève de MLP et SdA
Arnaud Bergeron <abergeron@gmail.com>
parents: 407
diff changeset
229 This is a common case where some specialized non-linearity is applied to the output layer only depending on the task at hand.
6330298791fb Description brève de MLP et SdA
Arnaud Bergeron <abergeron@gmail.com>
parents: 407
diff changeset
230
6330298791fb Description brève de MLP et SdA
Arnaud Bergeron <abergeron@gmail.com>
parents: 407
diff changeset
231 If you put 3 or more hidden layers in such a network you obtain what is called a deep MLP.
411
4f69d915d142 Better description of the model parameters.
Arnaud Bergeron <abergeron@gmail.com>
parents: 410
diff changeset
232 The parameters to adapt are the weight matrix and the bias vector for each layer.
410
6330298791fb Description brève de MLP et SdA
Arnaud Bergeron <abergeron@gmail.com>
parents: 407
diff changeset
233
379
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
234 \subsubsection{Stacked Denoising Auto-Encoders (SDAE)}
422
e7790db265b1 Basic text for section 3, add a bit more detail to section 4.2.2
Arnaud Bergeron <abergeron@gmail.com>
parents: 417
diff changeset
235 \label{SdA}
379
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
236
410
6330298791fb Description brève de MLP et SdA
Arnaud Bergeron <abergeron@gmail.com>
parents: 407
diff changeset
237 Auto-encoders are essentially a way to initialize the weights of the network to enable better generalization.
422
e7790db265b1 Basic text for section 3, add a bit more detail to section 4.2.2
Arnaud Bergeron <abergeron@gmail.com>
parents: 417
diff changeset
238 This is essentially unsupervised training where the layer is made to reconstruct its input through and encoding and decoding phase.
e7790db265b1 Basic text for section 3, add a bit more detail to section 4.2.2
Arnaud Bergeron <abergeron@gmail.com>
parents: 417
diff changeset
239 Denoising auto-encoders are a variant where the input is corrupted with random noise but the target is the uncorrupted input.
410
6330298791fb Description brève de MLP et SdA
Arnaud Bergeron <abergeron@gmail.com>
parents: 407
diff changeset
240 The principle behind these initialization methods is that the network will learn the inherent relation between portions of the data and be able to represent them thus helping with whatever task we want to perform.
6330298791fb Description brève de MLP et SdA
Arnaud Bergeron <abergeron@gmail.com>
parents: 407
diff changeset
241
411
4f69d915d142 Better description of the model parameters.
Arnaud Bergeron <abergeron@gmail.com>
parents: 410
diff changeset
242 An auto-encoder unit is formed of two MLP layers with the bottom one called the encoding layer and the top one the decoding layer.
4f69d915d142 Better description of the model parameters.
Arnaud Bergeron <abergeron@gmail.com>
parents: 410
diff changeset
243 Usually the top and bottom weight matrices are the transpose of each other and are fixed this way.
4f69d915d142 Better description of the model parameters.
Arnaud Bergeron <abergeron@gmail.com>
parents: 410
diff changeset
244 The network is trained as such and, when sufficiently trained, the MLP layer is initialized with the parameters of the encoding layer.
4f69d915d142 Better description of the model parameters.
Arnaud Bergeron <abergeron@gmail.com>
parents: 410
diff changeset
245 The other parameters are discarded.
4f69d915d142 Better description of the model parameters.
Arnaud Bergeron <abergeron@gmail.com>
parents: 410
diff changeset
246
410
6330298791fb Description brève de MLP et SdA
Arnaud Bergeron <abergeron@gmail.com>
parents: 407
diff changeset
247 The stacked version is an adaptation to deep MLPs where you initialize each layer with a denoising auto-encoder starting from the bottom.
411
4f69d915d142 Better description of the model parameters.
Arnaud Bergeron <abergeron@gmail.com>
parents: 410
diff changeset
248 During the initialization, which is usually called pre-training, the bottom layer is treated as if it were an isolated auto-encoder.
4f69d915d142 Better description of the model parameters.
Arnaud Bergeron <abergeron@gmail.com>
parents: 410
diff changeset
249 The second and following layers receive the same treatment except that they take as input the encoded version of the data that has gone through the layers before it.
410
6330298791fb Description brève de MLP et SdA
Arnaud Bergeron <abergeron@gmail.com>
parents: 407
diff changeset
250 For additional details see \cite{vincent:icml08}.
6330298791fb Description brève de MLP et SdA
Arnaud Bergeron <abergeron@gmail.com>
parents: 407
diff changeset
251
379
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
252 \section{Experimental Results}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
253
392
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
254 \subsection{SDA vs MLP}
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
255
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
256 \begin{center}
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
257 \begin{tabular}{lcc}
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
258 & train w/ & train w/ \\
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
259 & NIST & P07 + NIST \\ \hline
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
260 SDA & & \\ \hline
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
261 MLP & & \\ \hline
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
262 \end{tabular}
5f8fffd7347f possible image for illustrating perturbations
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents: 381
diff changeset
263 \end{center}
379
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
264
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
265 \subsection{Perturbed Training Data More Helpful for SDAE}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
266
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
267 \subsection{Training with More Classes than Necessary}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
268
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
269 \section{Conclusions}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
270
407
fe2e2964e7a3 description des transformations en cours ajout d un fichier special.bib pour des references specifiques
Xavier Glorot <glorotxa@iro.umontreal.ca>
parents: 393
diff changeset
271 \bibliography{strings,ml,aigaion,specials}
379
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
272 \bibliographystyle{mlapa}
a21a174c1c18 added writeup skeleton
Yoshua Bengio <bengioy@iro.umontreal.ca>
parents:
diff changeset
273
407
fe2e2964e7a3 description des transformations en cours ajout d un fichier special.bib pour des references specifiques
Xavier Glorot <glorotxa@iro.umontreal.ca>
parents: 393
diff changeset
274 \end{document}