p03 neural networks cvpr2012 deep learning methods for vision
TRANSCRIPT
![Page 1: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/1.jpg)
NEURAL NETS FOR VISIONCVPR 2012 Tutorial on Deep Learning
Part III
Marc'Aurelio Ranzato [email protected]/~ranzato
![Page 2: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/2.jpg)
2
Building an Object Recognition System
“CAR”
CLASSIFIERFEATURE
EXTRACTOR
IDEA: Use data to optimize features for the given task.
Ranzato
![Page 3: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/3.jpg)
3
Building an Object Recognition System
“CAR”
CLASSIFIER
What we want: Use parameterized function such that a) features are computed efficiently b) features can be trained efficiently
Ranzato
f X ;
![Page 4: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/4.jpg)
4
Building an Object Recognition System
“CAR”END-TO-END
RECOGNITIONSYSTEM
– Everything becomes adaptive. – No distiction between feature extractor and classifier.– Big non-linear system trained from raw pixels to labels.
Ranzato
![Page 5: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/5.jpg)
5
Building an Object Recognition System
“CAR”END-TO-END
RECOGNITIONSYSTEM
Q: How can we build such a highly non-linear system?
Ranzato
A: By combining simple building blocks we can make more and more complex systems.
![Page 6: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/6.jpg)
6
Building A Complicated Function
Ranzato
sin x
cos x
log x
exp x
x3log cos exp sin3x
Simple Functions
One Example of Complicated Function
![Page 7: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/7.jpg)
7
Building A Complicated Function
Ranzato
sin x
cos x
log x
exp x
x3log cos exp sin3x
Simple Functions
– Function composition is at the core of deep learning methods. – Each “simple function” will have parameters subject to training.
One Example of Complicated Function
![Page 8: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/8.jpg)
8
Implementing A Complicated Function
Ranzato
log cos exp sin3x
Complicated Function
sin x x3 exp x cos x log x
![Page 9: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/9.jpg)
9
Intuition Behind Deep Neural Nets
“CAR”
Ranzato
![Page 10: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/10.jpg)
10
Intuition Behind Deep Neural Nets
“CAR”
Ranzato
NOTE: Each black box can have trainable parameters. Their composition makes a highly non-linear system.
![Page 11: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/11.jpg)
11
Intuition Behind Deep Neural Nets
“CAR”
Ranzato
NOTE: System produces a hierarchy of features.
Intermediate representations/features
![Page 12: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/12.jpg)
12
Ranzato
“CAR”
Lee et al. “Convolutional DBN's for scalable unsup. learning...” ICML 2009
Intuition Behind Deep Neural Nets
![Page 13: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/13.jpg)
13
Ranzato
“CAR”
Lee et al. ICML 2009
Intuition Behind Deep Neural Nets
![Page 14: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/14.jpg)
14
Ranzato
“CAR”
Lee et al. ICML 2009
Intuition Behind Deep Neural Nets
![Page 15: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/15.jpg)
15
IDEA # 1Learn features from data
IDEA # 2Use differentiable functions that produce
features efficiently
IDEA # 3End-to-end learning:
no distinction between feature extractor and classifier
IDEA # 4“Deep” architectures:
cascade of simpler non-linear modules
KEY IDEAS OF NEURAL NETS
Ranzato
![Page 16: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/16.jpg)
16
KEY QUESTIONS
- What is the input-output mapping?
- How are parameters trained?
- How computational expensive is it?
- How well does it work?
Ranzato
![Page 17: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/17.jpg)
17
Outline- Neural Networks for Supervised Training
- Architecture- Loss function
- Neural Networks for Vision: Convolutional & Tiled- Unsupervised Training of Neural Networks- Extensions:
- semi-supervised / multi-task / multi-modal
- Comparison to Other Methods- boosting & cascade methods- probabilistic models
- Large-Scale Learning with Deep Neural Nets
Ranzato
![Page 18: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/18.jpg)
18
Outline- Neural Networks for Supervised Training
- Architecture- Loss function
- Neural Networks for Vision: Convolutional & Tiled- Unsupervised Training of Neural Networks- Extensions:
- semi-supervised / multi-task / multi-modal
- Comparison to Other Methods- boosting & cascade methods- probabilistic models
- Large-Scale Learning with Deep Neural Nets
Ranzato
![Page 19: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/19.jpg)
19
Linear Classifier: SVMInput:
Binary label:
Parameters:
Output prediction:
Loss:
X ∈RD
y
W ∈RD
W T X
Ranzato
L=12∥W∥
2max [0,1−W
TX y ]
L
W T X y
Hinge Loss
1
![Page 20: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/20.jpg)
20
Linear Classifier: Logistic Regression
Ranzato
L=12∥W∥
2 log 1exp −W
TX y
L
W T X y
Log Loss
X ∈RD
y
W ∈RD
W T X
Input:
Binary label:
Parameters:
Output prediction:
Loss:
![Page 21: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/21.jpg)
21
Logistic Regression: Probabilistic Interpretation
p y=1∣X =1
1e−WT X
Ranzato
L=− log p y∣X
Q: What is the gradient of w.r.t. ?L W
Input:
Binary label:
Parameters:
Output prediction:
Loss:
X ∈RD
y
W ∈RD
W T X
1
![Page 22: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/22.jpg)
22
Logistic Regression: Probabilistic Interpretation
p y=1∣X =1
1e−WT X
Ranzato
L= log 1exp −W T X y
Q: What is the gradient of w.r.t. ?L W
Input:
Binary label:
Parameters:
Output prediction:
Loss:
X ∈RD
y
W ∈RD
W T X
1
![Page 23: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/23.jpg)
23
Ranzato
sin x
cos x
log x
exp x
x3−log
1
1e−WT X
Simple Functions
Complicated Function
![Page 24: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/24.jpg)
24
Logistic Regression: Computing Loss
Ranzato
−log 1
1e−WT X
Complicated Function
W T X1
1e−u−log p
Lu p
![Page 25: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/25.jpg)
25
Chain Rule
Ranzato
dLdx
x y
Given and ,
What is ?
y x dL /dy
dL /dx
dLdy
![Page 26: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/26.jpg)
26
Chain Rule
Ranzato
dLdx
dLdy
x y
dLdx
=dLdy
⋅dydx
Given and ,
What is ?
y x dL /dy
dL /dx
![Page 27: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/27.jpg)
27
Chain Rule
Ranzato
dLdx
dLdy
x y
All needed information is local!
dLdx
=dLdy
⋅dydx
Given and ,
What is ?
y x dL /dy
dL /dx
![Page 28: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/28.jpg)
28
Ranzato
Logistic Regression: Computing Gradients
Ranzato
W T X1
1e−u−log p
Lu p
dLdp
X
−1p
![Page 29: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/29.jpg)
29
Ranzato
Logistic Regression: Computing Gradients
Ranzato
W T X1
1e−u−log p
Lu p
dpdu
dLdp
X
p1− p −1p
![Page 30: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/30.jpg)
30
Ranzato
Logistic Regression: Computing Gradients
Ranzato
W T X1
1e−u−log p
Lu p
dudW
dpdu
dLdp
dLdW
=dLdp
⋅dpdu
⋅dudW
= p−1 X
X
−1p
p1− pX
![Page 31: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/31.jpg)
31
Ranzato
What Did We Learn?
Ranzato
LogisticRegression
- Logistic Regression- How to compute gradients of complicated functions
![Page 32: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/32.jpg)
32
Ranzato
LogisticRegression
LogisticRegression
Neural Network
![Page 33: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/33.jpg)
33
Neural Network
Ranzato
– A neural net can be thought of as a stack of logistic regression classifiers. Each input is the output of the previous layer.
LogisticRegression
LogisticRegression
LogisticRegression
NOTE: intermediate units can be thought of as linear classifiers trained with implicit target values.
![Page 34: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/34.jpg)
34
Key Computations: F-Prop / B-Prop
F-PROP
X Z
Ranzato
![Page 35: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/35.jpg)
35
{∂Z∂ X
,∂Z∂
}
∂ L∂ X
∂ L∂Z
∂L∂
Ranzato
B-PROP
Key Computations: F-Prop / B-Prop
![Page 36: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/36.jpg)
36
Neural Net: Training
Ranzato
Layer 3Layer 2Layer 1
A) Compute loss on small mini-batch
![Page 37: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/37.jpg)
37
Neural Net: Training
Ranzato
Layer 3Layer 2Layer 1
F-PROP
A) Compute loss on small mini-batch
![Page 38: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/38.jpg)
38
Neural Net: Training
Ranzato
Layer 3Layer 2Layer 1
F-PROP
A) Compute loss on small mini-batch
![Page 39: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/39.jpg)
39
Neural Net: Training
Ranzato
Layer 3Layer 2Layer 1
F-PROP
A) Compute loss on small mini-batch
![Page 40: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/40.jpg)
40
Neural Net: Training
Ranzato
Layer 3Layer 2Layer 1
A) Compute loss B) Compute gradient w.r.t. parameters
B-PROP
A) Compute loss on small mini-batch
![Page 41: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/41.jpg)
41
Neural Net: Training
Ranzato
Layer 3Layer 2Layer 1
B-PROP
A) Compute loss B) Compute gradient w.r.t. parametersA) Compute loss on small mini-batch
![Page 42: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/42.jpg)
42
Neural Net: Training
Ranzato
Layer 3Layer 2Layer 1
B-PROP
A) Compute loss B) Compute gradient w.r.t. parametersA) Compute loss on small mini-batch
![Page 43: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/43.jpg)
43
Neural Net: Training
Ranzato
Layer 3Layer 2Layer 1
A) Compute loss B) Compute gradient w.r.t. parametersC) Use gradient to update parameters −
dLd
A) Compute loss on small mini-batch
![Page 44: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/44.jpg)
44
NEURAL NET: ARCHITECTURE
Ranzato
W j∈RM×N , b j∈R
N
h j h j1h j1=W j1
T h jb j1
h j∈RM , h j1∈R
N
![Page 45: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/45.jpg)
45
Ranzato
h j h j1h j1=W j1
T h jb j1
x =1
1e−x
NEURAL NET: ARCHITECTURE
x
x
![Page 46: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/46.jpg)
46
Ranzato
NEURAL NET: ARCHITECTURE
x = tanh x
h j h j1h j1=W j1
T h jb j1
x
x
![Page 47: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/47.jpg)
47
is equivalent to
Ranzato
Graphical Notations
f X ;W
W
X h
hk is called feature, hidden unit, neuron or code unit
![Page 48: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/48.jpg)
48
MOST COMMON ARCHITECTURE
yX
Errory
NOTE: Multi-layer neural nets with more than two layers are nowadays called deep nets!!
Ranzato
![Page 49: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/49.jpg)
49
NOTE: User must specify number of layers, number of hidden units, type of layers and loss function.
Ranzato
MOST COMMON ARCHITECTURE
yX
Errory
NOTE: Multi-layer neural nets with more than two layers are nowadays called deep nets!!
![Page 50: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/50.jpg)
50
MOST COMMON LOSSES
L=12∑i=1
N y i− y i
2
Square Euclidean Distance (regression):
Ranzato
yX
Errory
y , y∈RN
![Page 51: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/51.jpg)
51
Cross Entropy (classification):
L=−∑i=1
Ny i log y i
Ranzato
MOST COMMON LOSSES
yX
Errory
y , y∈[0,1]N , ∑i=1
Ny i=1, ∑i=1
Ny i=1
![Page 52: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/52.jpg)
52
1: User specifies loss based on the task.
Ranzato
NEURAL NETS FACTS
2: Any optimization algorithm can be chosen for training.
3: Cost of F-Prop and B-Prop is similar and proportional to the number of layers and their size.
![Page 53: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/53.jpg)
53
Toy Code: Neural Net Trainer% F-PROPfor i = 1 : nr_layers - 1 [h{i} jac{i}] = logistic(W{i} * h{i-1} + b{i});endh{nr_layers-1} = W{nr_layers-1} * h{nr_layers-2} + b{nr_layers-1};prediction = softmax(h{l-1});
% CROSS ENTROPY LOSSloss = - sum(sum(log(prediction) .* target));
% B-PROPdh{l-1} = prediction - target;for i = nr_layers – 1 : -1 : 1 Wgrad{i} = dh{i} * h{i-1}'; bgrad{i} = sum(dh{i}, 2); dh{i-1} = (W{i}' * dh{i}) .* jac{i-1}; end
% UPDATEfor i = 1 : nr_layers - 1 W{i} = W{i} – (lr / batch_size) * Wgrad{i}; b{i} = b{i} – (lr / batch_size) * bgrad{i}; end Ranzato
![Page 54: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/54.jpg)
54
TOY EXAMPLE: SYNTHETIC DATA1 input & 1 output100 hidden units in each layer
Ranzato
![Page 55: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/55.jpg)
55
1 input & 1 output3 hidden layers
Ranzato
TOY EXAMPLE: SYNTHETIC DATA
![Page 56: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/56.jpg)
56
1 input & 1 output3 hidden layers, 1000 hiddens Regression of cosine
Ranzato
TOY EXAMPLE: SYNTHETIC DATA
![Page 57: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/57.jpg)
57
1 input & 1 output3 hidden layers, 1000 hiddens Regression of cosine
Ranzato
TOY EXAMPLE: SYNTHETIC DATA
![Page 58: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/58.jpg)
58
Outline- Neural Networks for Supervised Training
- Architecture- Loss function
- Neural Networks for Vision: Convolutional & Tiled- Unsupervised Training of Neural Networks- Extensions:
- semi-supervised / multi-task / multi-modal
- Comparison to Other Methods- boosting & cascade methods- probabilistic models
- Large-Scale Learning with Deep Neural Nets
Ranzato
![Page 59: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/59.jpg)
59
Example: 1000x1000 image 1M hidden units
10^12 parameters!!!
- Spatial correlation is local- Better to put resources elsewhere!
Ranzato
FULLY CONNECTED NEURAL NET
![Page 60: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/60.jpg)
60
LOCALLY CONNECTED NEURAL NET
Ranzato
Example: 1000x1000 image 1M hidden units Filter size: 10x10
100M parameters
![Page 61: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/61.jpg)
61
LOCALLY CONNECTED NEURAL NET
Ranzato
Example: 1000x1000 image 1M hidden units Filter size: 10x10
100M parameters
![Page 62: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/62.jpg)
62
LOCALLY CONNECTED NEURAL NET
Ranzato
STATIONARITY? Statistics is similar at different locations
Example: 1000x1000 image 1M hidden units Filter size: 10x10
100M parameters
![Page 63: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/63.jpg)
63
CONVOLUTIONAL NET
Share the same parameters across different locations: Convolutions with learned kernels
Ranzato
![Page 64: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/64.jpg)
64
Learn multiple filters.
E.g.: 1000x1000 image 100 Filters Filter size: 10x10
10K parameters
Ranzato
CONVOLUTIONAL NET
![Page 65: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/65.jpg)
65
NEURAL NETS FOR VISION
A standard neural net applied to images:- scales quadratically with the size of the input- does not leverage stationarity
Solution:- connect each hidden unit to a small patch of the input- share the weight across hidden unitsThis is called: convolutional network.
LeCun et al. “Gradient-based learning applied to document recognition” IEEE 1998
Ranzato
![Page 66: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/66.jpg)
66
Let us assume filter is an “eye” detector.
Q.: how can we make the detection robust to the exact location of the eye?
Ranzato
CONVOLUTIONAL NET
![Page 67: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/67.jpg)
67
By “pooling” (e.g., max or average) filterresponses at different locations we gainrobustness to the exact spatial locationof features.
Ranzato
CONVOLUTIONAL NET
![Page 68: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/68.jpg)
68
CONV NETS: EXTENSIONSOver the years, some new modules have proven to be very effective when plugged into conv-nets:
- L2 Pooling
- Local Contrast Normalization
h i1, x , y=∑ j , k ∈N x , y h i , j , k2
h i1, x , y=hi , x , y−mi , N x , y
i , N x , y
layer i1layer i
x , yN x , y
layer i1layer i
x , yN x , y
Jarrett et al. “What is the best multi-stage architecture for object recognition?” ICCV 2009 Ranzato
![Page 69: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/69.jpg)
69
CONV NETS: L2 POOLING
-1 0 0 0 0
+1
Kavukguoglu et al. “Learning invariant features ...” CVPR 2009 Ranzato
h i h i1∑i=1
5⋅i
2
L2 Pooling
![Page 70: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/70.jpg)
70
CONV NETS: L2 POOLING
0 0 0 0+1
+1
Kavukguoglu et al. “Learning invariant features ...” CVPR 2009 Ranzato
h i h i1∑i=1
5⋅i
2
L2 Pooling
![Page 71: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/71.jpg)
71
LOCAL CONTRAST NORMALIZATION
h i1, x , y=hi , x , y−mi , N x , y
i , N x , y
0
1
0
-1
0
0
0.5
0
-0.5
0
LCN
Ranzato
![Page 72: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/72.jpg)
72
1
11
1
-9
1
LCN
0
0.5
0
-0.5
0
Ranzato
h i1, x , y=hi , x , y−mi , N x , y
i , N x , y
LOCAL CONTRAST NORMALIZATION
![Page 73: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/73.jpg)
73
L2 Pooling & Local Contrast Normalizationhelp learning more invariant representations!
CONV NETS: EXTENSIONS
Ranzato
![Page 74: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/74.jpg)
74
CONV NETS: TYPICAL ARCHITECTURE
Filtering Pooling LCN
One stage (zoom)
LinearLayer
Whole system
1st stage 2nd stage 3rd stage
Input Image
ClassLabels
Ranzato
![Page 75: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/75.jpg)
75
CONV NETS: TRAINING
Algorithm:Given a small mini-batch- FPROP- BPROP- PARAMETER UPDATE
Since convolutions and sub-sampling are differentiable, we can use standard back-propagation.
Ranzato
![Page 76: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/76.jpg)
76
CONV NETS: EXAMPLES
- Object category recognition Boureau et al. “Ask the locals: multi-way local pooling for image recognition” ICCV 2011- Segmentation Turaga et al. “Maximin learning of image segmentation” NIPS 2009- OCR Ciresan et al. “MCDNN for Image Classification” CVPR 2012- Pedestrian detection Kavukcuoglu et al. “Learning convolutional feature hierarchies for visual recognition” NIPS 2010- Robotics Sermanet et al. “Mapping and planning..with long range perception” IROS 2008
Ranzato
![Page 77: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/77.jpg)
77
LIMITATIONS & SOLUTIONS
- requires lots of labeled data to train
- difficult optimization
- scalability
Ranzato
+ unsupervised learning
+ layer-wise training
+ distributed training
![Page 78: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/78.jpg)
78
Outline- Neural Networks for Supervised Training
- Architecture- Loss function
- Neural Networks for Vision: Convolutional & Tiled- Unsupervised Training of Neural Networks- Extensions:
- semi-supervised / multi-task / multi-modal
- Comparison to Other Methods- boosting & cascade methods- probabilistic models
- Large-Scale Learning with Deep Neural Nets
Ranzato
![Page 79: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/79.jpg)
79
BACK TO LOGISTIC REGRESSION
Ranzato
Error
input
target
prediction
![Page 80: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/80.jpg)
80
Unsupervised Learning
Ranzato
Error
input prediction
?
![Page 81: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/81.jpg)
81
Unsupervised Learning
Ranzato
Q: How should we train the input-output mapping if we do not have target values?
A: Code has to retain information from the input but only if this is similar to training samples.
By better representing only those inputs that are similar to training samples we hope to extract interesting structure (e.g., structure of manifold where data live).
input code
![Page 82: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/82.jpg)
82
Unsupervised Learning
Ranzato
Q: How to constrain the model to represent training samples better than other data points?
![Page 83: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/83.jpg)
83
Unsupervised Learning
Ranzato
– reconstruct the input from the code & make code compact (auto-encder with bottle-neck).
– reconstruct the input from the code & make code sparse (sparse auto-encoders)
– add noise to the input or code (denoising auto-encoders)
– make sure that the model defines a distribution that normalizes to 1 (RBM).
see work in LeCun, Ng, Fergus, Lee, Yu's labs
see work in Y. Bengio, Lee's lab
see work in Y. Bengio, Hinton, Lee, Salakthudinov's lab
![Page 84: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/84.jpg)
84
AUTO-ENCODERS NEURAL NETS
Ranzato
encoder
Error
input reconstructiondecoder
code
– input higher dimensional than code - error: ||reconstruction - input|| - training: back-propagation
2
![Page 85: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/85.jpg)
85
SPARSE AUTO-ENCODERS
Ranzato
encoder
Error
input
code
– sparsity penalty: ||code||- error: ||reconstruction - input||
- training: back-propagation
SparsityPenalty
1
- loss: sum of squared reconstruction error and sparsity
decoderreconstruction
2
![Page 86: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/86.jpg)
86
SPARSE AUTO-ENCODERS
Ranzato
encoder
Error
input
code
– input: code:
- loss:
SparsityPenalty
Le et al. “ICA with reconstruction cost..” NIPS 2011
h=W T XX
L X ;W =∥W h−X∥2∑ j∣h j∣
decoderreconstruction
![Page 87: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/87.jpg)
87
How To Use Unsupervised Learning
Ranzato
1) Given unlabeled data, learn features
![Page 88: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/88.jpg)
88
How To Use Unsupervised Learning
Ranzato
1) Given unlabeled data, learn features2) Use encoder to produce features and train another layer on the top
![Page 89: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/89.jpg)
89
How To Use Unsupervised Learning
Ranzato
Layer-wise training of a feature hierarchy
1) Given unlabeled data, learn features2) Use encoder to produce features and train another layer on the top
![Page 90: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/90.jpg)
90
How To Use Unsupervised Learning
Ranzato
1) Given unlabeled data, learn features2) Use encoder to produce features and train another layer on the top3) feed features to classifier & train just the classifier
Reduced overfitting since features are learned in unsupervised way!
input
label
prediction
![Page 91: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/91.jpg)
91
How To Use Unsupervised Learning
Ranzato
1) Given unlabeled data, learn features2) Use encoder to produce features and train another layer on the top3) feed features to classifier & jointly train the whole system
Given enough data, this usually yields the best results: end-to-end learning!
input
label
prediction
![Page 92: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/92.jpg)
92
Outline- Neural Networks for Supervised Training
- Architecture- Loss function
- Neural Networks for Vision: Convolutional & Tiled- Unsupervised Training of Neural Networks- Extensions:
- semi-supervised / multi-task / multi-modal
- Comparison to Other Methods- boosting & cascade methods- probabilistic models
- Large-Scale Learning with Deep Neural Nets
Ranzato
![Page 93: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/93.jpg)
93
Semi-Supervised Learning
Ranzato
airplane
truck
deer
frog
bird
![Page 94: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/94.jpg)
94
Semi-Supervised Learning
Ranzato
airplane
truck
deer
frog
bird
LOTS & LOTS OF UNLABELED DATA!!!
![Page 95: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/95.jpg)
95
Ranzato
Semi-Supervised Learning
input
prediction
Weston et al. “Deep learning via semi-supervised embedding” ICML 2008
label
Loss = supervised_error + unsupervised_error
![Page 96: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/96.jpg)
96
Multi-Task Learning
Ranzato
Face detection is hard because of lighting, pose, but also occluding goggles.
Face detection could made be easier by face identification.
The identification task may help the detection task.
![Page 97: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/97.jpg)
97
Multi-Task Learning- Easy to add many error terms to loss function.- Joint learning of related tasks yields better representations.
Example of architecture:
Collobert et al. “NLP (almost) from scratch” JMLR 2011 Ranzato
![Page 98: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/98.jpg)
98
Multi-Modal Learning
Ranzato
Audio and Video streams are often complimentary to each other.
E.g., audio can provide important clues to improve visual recognition, and vice versa.
![Page 99: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/99.jpg)
99
Multi-Modal Learning
- Weak assumptions on input distribution- Fully adaptive to data
Ngiam et al. “Multi-modal deep learning” ICML 2011 Ranzato
Example of architecture:modality #1
modality #2
modality #3
![Page 100: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/100.jpg)
100
Outline- Neural Networks for Supervised Training
- Architecture- Loss function
- Neural Networks for Vision: Convolutional & Tiled- Unsupervised Training of Neural Networks- Extensions:
- semi-supervised / multi-task / multi-modal
- Comparison to Other Methods- boosting & cascade methods- probabilistic models
- Large-Scale Learning with Deep Neural Nets
Ranzato
![Page 101: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/101.jpg)
101
Boosting & Forests
Deep Nets:- single highly non-linear system- “deep” stack of simpler modules- all parameters are subject to learning
Boosting & Forests:- sequence of “weak” (simple) classifiers that are linearly combined to produce a powerful classifier- subsequent classifiers do not exploit representations of earlier classifiers, it's a “shallow” linear mixture
- typically features are not learned Ranzato
![Page 102: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/102.jpg)
102
Properties Deep Nets Boosting
Hierarchical features
Easy to parallelize
End-to-end learning
Fast training
Fast at test time
Leverage unlab. data
Adaptive features
Ranzato
![Page 103: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/103.jpg)
103
Deep Neural-Nets VS Probabilistic ModelsDeep Neural Nets:- mean-field approximations of intractable probabilistic models- usually more efficient- typically more unconstrained (partition function has to be replaced by other constraints, e.g. sparsity).
Hierarchical Probabilistic Models (DBN, DBM, etc.):- in the most interesting cases, they are intractable- they better deal with uncertainty- they can be easily combined
Ranzato
![Page 104: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/104.jpg)
104
Example: Auto-Encoder
E [Z∣X ]= W T Xbe
Neural Net:
Probabilistic Model (Gaussian RBM):
X=W d Zbd
Z= W eT Xbe
E [ X ∣Z ]=W Zbd
Ranzato
code
reconstruction
![Page 105: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/105.jpg)
105
Properties Deep Nets Probab. Models
Hierarchical features
Models uncertainty
End-to-end learning
Fast training
Fast at test time
Leverage unlab. data
Adaptive features
Ranzato
![Page 106: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/106.jpg)
106
Outline- Neural Networks for Supervised Training
- Architecture- Loss function
- Neural Networks for Vision: Convolutional & Tiled- Unsupervised Training of Neural Networks- Extensions:
- semi-supervised / multi-task / multi-modal
- Comparison to Other Methods- boosting & cascade methods- probabilistic models
- Large-Scale Learning with Deep Neural Nets
Ranzato
![Page 107: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/107.jpg)
107
Tera-Scale Deep Learning @ Google
Observation #1: more features always improve performance unless data is scarce.
Observation #2: deep learning methods have higher capacity and have the potential to model data better.
Q #1: Given lots of data and lots of machines, can we scale up deep learning methods?
Q #2: Will deep learning methods perform much better?
Ranzato
![Page 108: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/108.jpg)
108
The Challenge
– best optimizer in practice is on-line SGD which is naturally sequential, hard to parallelize.
– layers cannot be trained independently and in parallel, hard to distribute
– model can have lots of parameters that may clog the network, hard to distribute across machines
A Large Scale problem has: – lots of training samples (>10M)– lots of classes (>10K) and – lots of input dimensions (>10K).
Ranzato
![Page 109: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/109.jpg)
109
Our Solution
input
1st layer
2nd layer
Le et al. “Building high-level features using large-scale unsupervised learning” ICML 2012 Ranzato
![Page 110: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/110.jpg)
110
1st machine
2nd
machine3rd
machine
Ranzato
Our Solution
input
1st layer
2nd layer
![Page 111: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/111.jpg)
111
MODEL PARALLELISM
Ranzato
Our Solution
1st machine
2nd
machine3rd
machine
![Page 112: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/112.jpg)
112
Distributed Deep Nets
Deep Net MODEL PARALLELISM
input #1
RanzatoLe et al. “Building high-level features using large-scale unsupervised learning” ICML 2012
![Page 113: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/113.jpg)
113
Distributed Deep Nets
MODEL PARALLELISM
+DATA
PARALLELISM
RanzatoLe et al. “Building high-level features using large-scale unsupervised learning” ICML 2012
input #1input #2
input #3
![Page 114: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/114.jpg)
114
Asynchronous SGD
PARAMETER SERVER
1st replica 2nd replica 3rd replicaRanzato
![Page 115: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/115.jpg)
115
Asynchronous SGD
PARAMETER SERVER
1st replica 2nd replica 3rd replicaRanzato
∂ L∂1
![Page 116: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/116.jpg)
116
Asynchronous SGD
PARAMETER SERVER
1st replica 2nd replica 3rd replicaRanzato
1
![Page 117: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/117.jpg)
117
Asynchronous SGD
PARAMETER SERVER(update parameters)
1st replica 2nd replica 3rd replicaRanzato
![Page 118: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/118.jpg)
118
Asynchronous SGD
PARAMETER SERVER
1st replica 2nd replica 3rd replicaRanzato
∂ L∂2
![Page 119: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/119.jpg)
119
Asynchronous SGD
PARAMETER SERVER
1st replica 2nd replica 3rd replicaRanzato
2
![Page 120: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/120.jpg)
120
Asynchronous SGD
PARAMETER SERVER(update parameters)
1st replica 2nd replica 3rd replicaRanzato
![Page 121: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/121.jpg)
121
Unsupervised Learning With 1B Parameters
DATA: 10M youtube (unlabeled) frames of size 200x200.
RanzatoLe et al. “Building high-level features using large-scale unsupervised learning” ICML 2012
![Page 122: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/122.jpg)
122
Deep Net:– 3 stages
– each stage consists of local filtering, L2 pooling, LCN - 18x18 filters- 8 filters at each location- L2 pooling and LCN over 5x5 neighborhoods
– training jointly the three layers by:- reconstructing the input of each layer- sparsity on the code
RanzatoLe et al. “Building high-level features using large-scale unsupervised learning” ICML 2012
Unsupervised Learning With 1B Parameters
![Page 123: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/123.jpg)
123
Deep Net:– 3 stages
– each stage consists of local filtering, L2 pooling, LCN - 18x18 filters- 8 filters at each location- L2 pooling and LCN over 5x5 neighborhoods
– training jointly the three layers by:- reconstructing the input of each layer- sparsity on the code
RanzatoLe et al. “Building high-level features using large-scale unsupervised learning” ICML 2012
Unsupervised Learning With 1B Parameters
1B parameters!!!
![Page 124: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/124.jpg)
124
Validating Unsupervised Learning
RanzatoLe et al. “Building high-level features using large-scale unsupervised learning” ICML 2012
The network has seen lots of objects during training, but without any label.
Q.: how can we validate unsupervised learning?
Q.: Did the network form any high-level representation? E.g., does it have any neuron responding for faces?
– build validation set with 50% faces, 50% random images- study properties of neurons
![Page 125: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/125.jpg)
125
Validating Unsupervised Learning
1st stage 2nd stage 3rd stage
neu
ron
respo
nses
RanzatoLe et al. “Building high-level features using large-scale unsupervised learning” ICML 2012
![Page 126: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/126.jpg)
126
Top Images For Best Face Neuron
Ranzato
![Page 127: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/127.jpg)
127
Best Input For Face Neuron
RanzatoLe et al. “Building high-level features using large-scale unsupervised learning” ICML 2012
![Page 128: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/128.jpg)
128
Unsupervised + Supervised (ImageNet)
1st stage 2nd stage 3rd stage
Input Image y
COST
y
RanzatoLe et al. “Building high-level features using large-scale unsupervised learning” ICML 2012
![Page 129: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/129.jpg)
129
Object Recognition on ImageNet
IMAGENET v.2011 (16M images, 20K categories)
METHOD ACCURACY %
Weston & Bengio 2011
Deep Net (from random)
Deep Net (from unsup.)
9.3
13.6
15.8
Linear Classifier on deep features 13.1
RanzatoLe et al. “Building high-level features using large-scale unsupervised learning” ICML 2012
![Page 130: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/130.jpg)
130
Top Inputs After Supervision
Ranzato
![Page 131: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/131.jpg)
131
Ranzato
Top Inputs After Supervision
![Page 132: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/132.jpg)
132
Experiments: and many more...
- automatic speech recognition
- natural language processing
- biomed applications
- finance
Generic learning algorithm!!
Ranzato
![Page 133: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/133.jpg)
133
ReferencesTutorials & Background Material
Ranzato
– Yoshua Bengio, Learning Deep Architectures for AI, Foundations and Trends in Machine Learning, 2(1), pp.1-127, 2009.– LeCun, Chopra, Hadsell, Ranzato, Huang: A Tutorial on Energy-Based Learning, in Bakir, G. and Hofman, T. and Schölkopf, B. and Smola, A. and Taskar, B. (Eds), Predicting Structured Data, MIT Press, 2006
Convolutional Nets– LeCun, Bottou, Bengio and Haffner: Gradient-Based Learning Applied to Document Recognition, Proceedings of the IEEE, 86(11):2278-2324, November 1998– Jarrett, Kavukcuoglu, Ranzato, LeCun: What is the Best Multi-Stage Architecture for Object Recognition?, Proc. International Conference on Computer Vision (ICCV'09), IEEE, 2009- Kavukcuoglu, Sermanet, Boureau, Gregor, Mathieu, LeCun: Learning Convolutional Feature Hierachies for Visual Recognition, Advances in Neural Information Processing Systems (NIPS 2010), 23, 2010
![Page 134: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/134.jpg)
134
Unsupervised Learning
Ranzato
– ICA with Reconstruction Cost for Efficient Overcomplete Feature Learning. Le, Karpenko, Ngiam, Ng. In NIPS*2011– Rifai, Vincent, Muller, Glorot, Bengio, Contracting Auto-Encoders: Explicit invariance during feature extraction, in: Proceedings of the Twenty-eight International Conference on Machine Learning (ICML'11), 2011- Vincent, Larochelle, Lajoie, Bengio, Manzagol, Stacked Denoising Autoencoders: Learning Useful Representations in a Deep Network with a Local Denoising Criterion, Journal of Machine Learning Research, 11:3371--3408, 2010.- Gregor, Szlam, LeCun: Structured Sparse Coding via Lateral Inhibition, Advances in Neural Information Processing Systems (NIPS 2011), 24, 2011- Kavukcuoglu, Ranzato, LeCun. "Fast Inference in Sparse Coding Algorithms with Applications to Object Recognition". ArXiv 1010.3467 2008
- Hinton, Krizhevsky, Wang, Transforming Auto-encoders, ICANN, 2011
Multi-modal Learning– Multimodal deep learning, Ngiam, Khosla, Kim, Nam, Lee, Ng. In Proceedings of the Twenty-Eighth International Conference on Machine Learning, 2011.
![Page 135: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/135.jpg)
135
– Gregor, LeCun “Emergence of complex-like cells in a temporal product network with local receptive fields” Arxiv. 2009– Ranzato, Mnih, Hinton “Generating more realistic images using gated MRF's” NIPS 2010– Le, Ngiam, Chen, Chia, Koh, Ng “Tiled convolutional neural networks” NIPS 2010
Locally Connected Nets
Ranzato
Distributed Learning– Le, Ranzato, Monga, Devin, Corrado, Chen, Dean, Ng. "Building High-Level Features Using Large Scale Unsupervised Learning". International Conference of Machine Learning (ICML 2012), Edinburgh, 2012.
Papers on Scene Parsing– Farabet, Couprie, Najman, LeCun, “Scene Parsing with Multiscale Feature Learning, Purity Trees, and Optimal Covers”, in Proc. of the International Conference on Machine Learning (ICML'12), Edinburgh, Scotland, 2012.
– Farabet, Couprie, Najman, LeCun, “Scene Parsing with Multiscale Feature Learning, Purity Trees, and Optimal Covers”, in Proc. of the International Conference on Machine Learning (ICML'12), Edinburgh, Scotland, 2012.- Socher, Lin, Ng, Manning, “Parsing Natural Scenes and Natural Language with Recursive Neural Networks”. International Conference of Machine Learning (ICML 2011) 2011.
![Page 136: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/136.jpg)
136
Papers on Object Recognition
Ranzato
- Boureau, Le Roux, Bach, Ponce, LeCun: Ask the locals: multi-way local pooling for image recognition, Proc. International Conference on Computer Vision 2011- Sermanet, LeCun: Traffic Sign Recognition with Multi-Scale Convolutional Networks, Proceedings of International Joint Conference on Neural Networks (IJCNN'11)- Ciresan, Meier, Gambardella, Schmidhuber. Convolutional Neural Network Committees For Handwritten Character Classification. 11th International Conference on Document Analysis and Recognition (ICDAR 2011), Beijing, China.- Ciresan, Meier, Masci, Gambardella, Schmidhuber. Flexible, High Performance Convolutional Neural Networks for Image Classification. International Joint Conference on Artificial Intelligence IJCAI-2011.Papers on Action Recognition
– Learning hierarchical spatio-temporal features for action recognition with independent subspace analysis, Le, Zou, Yeung, Ng. In Computer Vision and Pattern Recognition (CVPR), 2011Papers on Segmentation– Turaga, Briggman, Helmstaedter, Denk, Seung Maximin learning of image segmentation. NIPS, 2009.
![Page 137: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/137.jpg)
137
Papers on Vision for Robotics
Ranzato
– Hadsell, Sermanet, Scoffier, Erkan, Kavackuoglu, Muller, LeCun: Learning Long-Range Vision for Autonomous Off-Road Driving, Journal of Field Robotics, 26(2):120-144, February 2009,
– Serre, Wolf, Bileschi, Riesenhuber, Poggio. Robust Object Recognition with Cortex-like Mechanisms, IEEE Transactions on Pattern Analysis and Machine Intelligence, 29, 3, 411-426, 2007.- Pinto, Doukhan, DiCarlo, Cox "A high-throughput screening approach to discovering good forms of biologically inspired visual representation." {PLoS} Computational Biology. 2009
Papers on Biological Inspired Vision
Deep Convex Nets & Deconv-Nets– Deng, Yu. “Deep Convex Network: A Scalable Architecture for Speech Pattern Classification.” Interspeech, 2011.- Zeiler, Taylor, Fergus "Adaptive Deconvolutional Networks for Mid and High Level Feature Learning." ICCV. 2011
![Page 138: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/138.jpg)
138
Papers on Embedded ConvNets for Real-Time Vision Applications
Ranzato
– Farabet, Martini, Corda, Akselrod, Culurciello, LeCun: NeuFlow: A Runtime Reconfigurable Dataflow Processor for Vision, Workshop on Embedded Computer Vision, CVPR 2011,
Papers on Image Denoising Using Neural Nets– Burger, Schuler, Harmeling: Image Denoisng: Can Plain Neural Networks Compete with BM3D?, Computer Vision and Pattern Recognition, CVPR 2012,
![Page 139: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/139.jpg)
139
Software & LinksDeep Learning website
Matlab code for R-ICA unsupervised algorithm
Python-based learning library
C++ code for ConvNets
Lush learning library which includes ConvNets
Torch7: learning library that supports neural net training
Ranzato
– http://deeplearning.net/
– http://eblearn.sourceforge.net/
– http://deeplearning.net/
- http://lush.sourceforge.net/
– http://eblearn.sourceforge.net/
- http://deeplearning.net/software/theano/
- http://ai.stanford.edu/~quocle/rica_release.zip
- http://lush.sourceforge.net/
http://www.torch.ch
![Page 140: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/140.jpg)
140
Software & Links
Code used to generate demo for this tutorial
Ranzato
- http://cs.nyu.edu/~fergus/tutorials/deep_learning_cvpr12/
![Page 141: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/141.jpg)
141
Acknowledgements
Ranzato
Quoc Le, Andrew Ng
Jeff Dean, Kai Chen, Greg Corrado, Matthieu Devin, Mark Mao, Rajat Monga, Paul Tucker, Samy Bengio
Yann LeCun, Pierre Sermanet, Clement Farabet
![Page 142: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/142.jpg)
142
Visualizing Learned Features
Ranzato et al. “Sparse feature learning for DBNs” NIPS 2007 Ranzato
Q: can we interpret the learned features?
![Page 143: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/143.jpg)
143
Visualizing Learned Features
Ranzato et al. “Sparse feature learning for DBNs” NIPS 2007 Ranzato
W h=W 1h1W 2h2reconstruction:
0.9 + 0.7 + 0.5 + 1.0 + ...=≈
Columns of show what each code unit represents.W
![Page 144: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/144.jpg)
144
Visualizing Learned Features
Ranzato et al. “Sparse feature learning for DBNs” NIPS 2007 Ranzato
1st layer features
![Page 145: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/145.jpg)
145
Visualizing Learned Features
Ranzato et al. “Sparse feature learning for DBNs” NIPS 2007 Ranzato
Q: How about the second layer features?
A: Similarly, each second layer code unit can be visualized by taking its bases and then projecting those bases in image space through the first layer decoder.
![Page 146: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/146.jpg)
146
Visualizing Learned Features
Ranzato et al. “Sparse feature learning for DBNs” NIPS 2007 Ranzato
Q: How about the second layer features?
Missing edges have 0 weight.Light gray nodes have zero value.
A: Similarly, each second layer code unit can be visualized by taking its bases and then projecting those bases in image space through the first layer decoder.
![Page 147: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/147.jpg)
147
Visualizing Learned Features
Ranzato et al. “Sparse feature learning for DBNs” NIPS 2007 Ranzato
1st layer features
2nd layer features
Q: how are these images computed?
![Page 148: P03 neural networks cvpr2012 deep learning methods for vision](https://reader034.vdocuments.mx/reader034/viewer/2022050808/554a0691b4c9055b7a8b55a0/html5/thumbnails/148.jpg)
148
Example of Feature Learning
Ranzato et al. “Sparse feature learning for DBNs” NIPS 2007
1st layer features
2nd layer features
Ranzato