SlideShare une entreprise Scribd logo
1  sur  52
Convolutional Restricted
Boltzmann Machines for
Feature Learning
Mohammad Norouzi
Advisor: Dr. Greg Mori
CS @ Simon Fraser University
27 Nov 2009 1
CRBMs for
Feature Learning
Mohammad Norouzi
Advisor: Dr. Greg Mori
CS @ Simon Fraser University
27 Nov 2009 2
Problems
Human detection
Handwritten digit
classification
3
Sliding Window Approach
4
Sliding Window Approach (Cont’d)
5
[INRIA Person Dataset]
Success or Failure of an object recognition
algorithm hinges on the features used
Input
Feature
representation
Label
Our Focus Classifier
? Human
Background
0 / 1 / 2 / 3 / …
6
Learning
Local Feature Detector Hierarchies
7
Larger More complicated Less frequent
Generative & Layerwise Learning
8
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
Generative
CRBM
?
?
? ?
?
?
?
?
? ?
?
?
Visual Features: Filtering
9
1 0 -1
2 0 -2
1 0 -1
Filter Kernel (Feature)
-1 0 1
-2 0 2
-1 0 1
0 -1 -2
1 0 -1
2 1 0
Filter Response
1
W
V
2
W 2
W
)
,
( 1
W
V
Filter )
,
( 2
W
V
Filter )
,
( 3
W
V
Filter
Our approach to feature learning
is generative
?
?
?
1
H
2
H
3
H
V
Binary Hidden
Variables
10
1
W
2
W
3
W
(CRBM model)
Related Work
11
Related Work
• Convolutional Neural Network (CNN)
– Filtering layers are bundled with a classifier, and all
the layers are learned together using error
backpropagation.
– Does not perform well on natural images
• Biologically plausible models
– Hand-crafted first layer vs. Randomly selected
prototypes for second layer.
[Lecun et al. 98]
[Ranzato et al. CVPR'07]
[Serre et al., PAMI'07] [Mutch and Lowe, CVPR'06]
12
Related Work (cont’d)
• Deep Belief Net
– A two layer partially observed MRF, called RBM, is
the building block
– Learning is performed unsupervised and layer-by-
layer from bottom layer upwards
• Our contributions: We incorporate spatial
locality into RBMs and adapt the learning
algorithm accordingly
• We add more complicated components such
as pooling and sparsity into deep belief nets
[Hinton et al., NC'2006]
13
Why Generative &Unsupervised
• Discriminative learning of deep and large
neural networks has not been successful
– Requires large training sets
– Easily gets over-fitted for large models
– First layer gradients are relatively small
• Alternative hybrid approach
– Learn a large set of first layer features generatively
– Switch to a discriminative model to select the
discriminative features from those that are learned
– Discriminative fine-tuning is helpful
Details
15
CRBM
• Image is the visible layer
and hidden layer is
related to filter responses
• An energy based
probabilistic model
16
Dot product of vectorized matrices
   
 
)
,
(
)
;
,
(
)
;
,
(
)
;
,
(
,
exp
1
k
k
k
k
k k
k
H
W
V
Filter
H
W
H
V
E
W
H
V
E
W
H
V
E
H;W
V
E
Z
=
V;W
P







Training CRBMs
• Maximum likelihood learning of CRBMs is difficult
• Contrastive Divergence (CD) learning is applicable
• For CD learning we need to compute the
conditionals and .
data
17
sample
 
H
V
P |  
V
H
P |
CRBM (Backward)
• Nearby hidden variables
cooperate in reconstruction
• Conditional Probabilities
take the form
18
 
 
)
exp
1
(
1
*
)
(
)
,
(
)
|
(
)
,
(
)
|
(
x
k k
k
k
k
x
W
H
Filter
H
V
P
W
V
Filter
V
H
P









Learning the Hierarchy
• The structure is trained bottom up and layerwise
• The CRBM model for training filtering layers
• Filtering layers are followed by down-sampling
CRBM CRBM
Classifier
Pooling Pooling
19
Filtering
Non-linearity
Reduce the
dimensionality
layers
Input
1st
Filters 2nd
Filters
Responses
Responses
1 3
2 4
Experiments
21
Evaluation
MNIST digit dataset
• Training set: 60,000 image
of digits of size 28x28
• Test set: 10,000 images
INRIA person dataset
• Training set: 2416 person
windows of size 128 x 64
pixels and 4.5x106 negative
windows
• Test set: 1132 positive and
2x106 negative windows
22
First layer filters
• Gray-scale images of
INRIA positive set
• 15 filters of 7x7
23
• MNIST unlabeled digits
• 15 filters of 5x5
Second Layer Features (MNIST)
• Hard to visualize the filters
• We show patches highly responded to filters:
24
24
Second Layer Features (INRIA)
25
MNIST Results
• MNIST error rate when model is trained on
the full training set
26
Results
27
False Positive
1st
28
2nd
29
3rd
30
4th
31
5th
32
INRIA Results
• Adding our large-scale features significantly
improves performance of the baseline (HOG)
33
Conclusion
• We extended the RBM model to Convolutional
RBM, useful for domains with spatial locality
• We exploited CRBMs to train local hierarchical
feature detectors one layer at a time and
generatively
• This method obtained results comparable to
state-of-the-art in digit classification and
human detection
34
Thank You 
35
Hierarchical Feature Detector
36
? ? ?
? ? ?
? ? ?
Contrastive Divergence Learning
37
 
data
1
k
data
0
k
k
k H
,
V
Filter
H
,
V
Filter
η
+
W
=
W )
(
)
( 1
0

   
k
k
H
V,
Filter
=
W
θ
H;
V,
E



Training CRBMs (Cont'd)
• The problem of reconstructing border region
becomes severe when number of Gibbs
sampling steps > 1.
– Partition visible units into middle and border
regions
• Instead of maximizing the
likelihood, we (approximately)
maximize  
 b
m
v
|
v
p
Enforcing Feature Sparsity
• The CRBM's representation is K (number of
filters) times overcomplete
• After a few CD learning iterations, V is
perfectly reconstructed
• Enforce sparsity to tackle this problem
– Hidden bias terms were frozen at large negative
values
• Having a single non-sparse hidden unit
improves the learned features
– Might be related to the ergodicity condition
Probabilistic Meaning of Max
1 2 3 4 5 6
1 2 3 4
Max
1 2 3 4 5 6
1 1 2 2
h
h'
v
 
6
4
5
3
4
2
3
1
:
T
4
:
T
3
:
T
2
:
T
1
v
w
h
+
v
w
h
+
v
w
h
+
v
w
h
=
h
v,
E

h'
v
   
 
6
4
5
3
4
2
3
1
:
T
2
:
T
2
:
T
1
:
T
1
v
w
h'
+
v
w
h'
max
+
v
w
h'
,
v
w
h'
max
=
h
v,
E

The Classifier Layer
• We used SVM as our final classifier
– RBF kernel for MNIST
– Linear kernel for INRIA
– For INRIA we combined our 4th layer outputs and
HOG features
• We experimentally observed that relaxing the
sparsity of CRBM's hidden units yields better
results
– This lets the discriminative model to set the
thresholds itself
Why HOG features are added?
• Because part-like features
are very sparse
• Having a template of the
human figure helps a lot
f
RBM
• Two layer pairwise MRF with a full set
of hidden-visible connections
• RBM Is an energy based model
• Hidden random variables are binary, Visible
variables can be binary or continuous
• Inference is straightforward: and
• Contrastive Divergence learning for training
h
v
w
 
 
 
 
θ
h;
v,
E
θ
Z
=
θ
h;
v,
p 
exp
1
  


 

 2
2
1
i
j
j
i
i
j
ij
i v
+
h
c
v
b
h
w
v
=
θ
h;
v,
E
 
v
|
h
p  
h
|
v
p
Why Unsupervised Bottom-Up
• Discriminative learning of deep structure has
not been successful
– Requires large training sets
– Easily is over-fitted for large models
– First layer gradients are relatively small
• Alternative hybrid approach
– Learn a large set of first layer features generatively
– Later, switch to a discriminative model to select
the discriminative features from those learned
– Fine-tune the features using
INRIA Results (Cont'd)
• Missrate at different FPPW rates
• FPPI is a better indicator of performance
• More experiments on size of features and
number of layers are desired
convolutional_rbm.ppt
convolutional_rbm.ppt
convolutional_rbm.ppt
convolutional_rbm.ppt
convolutional_rbm.ppt
convolutional_rbm.ppt
convolutional_rbm.ppt

Contenu connexe

Similaire à convolutional_rbm.ppt

Deep learning from a novice perspective
Deep learning from a novice perspectiveDeep learning from a novice perspective
Deep learning from a novice perspectiveAnirban Santara
 
Multi-class Classification on Riemannian Manifolds for Video Surveillance
Multi-class Classification on Riemannian Manifolds for Video SurveillanceMulti-class Classification on Riemannian Manifolds for Video Surveillance
Multi-class Classification on Riemannian Manifolds for Video SurveillanceDiego Tosato
 
CTF: Anomaly Detection in High-Dimensional Time Series with Coarse-to-Fine Mo...
CTF: Anomaly Detection in High-Dimensional Time Series with Coarse-to-Fine Mo...CTF: Anomaly Detection in High-Dimensional Time Series with Coarse-to-Fine Mo...
CTF: Anomaly Detection in High-Dimensional Time Series with Coarse-to-Fine Mo...ssuser9357dd
 
Comparison of Fine-tuning and Extension Strategies for Deep Convolutional Neu...
Comparison of Fine-tuning and Extension Strategies for Deep Convolutional Neu...Comparison of Fine-tuning and Extension Strategies for Deep Convolutional Neu...
Comparison of Fine-tuning and Extension Strategies for Deep Convolutional Neu...InVID Project
 
lecture15-neural-nets (2).pptx
lecture15-neural-nets (2).pptxlecture15-neural-nets (2).pptx
lecture15-neural-nets (2).pptxanjithaba
 
A Re-evaluation of Pedestrian Detection on Riemannian Manifolds
A Re-evaluation of Pedestrian Detection on Riemannian ManifoldsA Re-evaluation of Pedestrian Detection on Riemannian Manifolds
A Re-evaluation of Pedestrian Detection on Riemannian ManifoldsDiego Tosato
 
Convolutional Neural Networks (CNN)
Convolutional Neural Networks (CNN)Convolutional Neural Networks (CNN)
Convolutional Neural Networks (CNN)Gaurav Mittal
 
Content-Based Image Retrieval (D2L6 Insight@DCU Machine Learning Workshop 2017)
Content-Based Image Retrieval (D2L6 Insight@DCU Machine Learning Workshop 2017)Content-Based Image Retrieval (D2L6 Insight@DCU Machine Learning Workshop 2017)
Content-Based Image Retrieval (D2L6 Insight@DCU Machine Learning Workshop 2017)Universitat Politècnica de Catalunya
 
Leveraging high level and low-level features for multimedia event detection.2...
Leveraging high level and low-level features for multimedia event detection.2...Leveraging high level and low-level features for multimedia event detection.2...
Leveraging high level and low-level features for multimedia event detection.2...Lu Jiang
 
Fundamental of deep learning
Fundamental of deep learningFundamental of deep learning
Fundamental of deep learningStanley Wang
 
General Tips for participating Kaggle Competitions
General Tips for participating Kaggle CompetitionsGeneral Tips for participating Kaggle Competitions
General Tips for participating Kaggle CompetitionsMark Peng
 
Online Stochastic Tensor Decomposition for Background Subtraction in Multispe...
Online Stochastic Tensor Decomposition for Background Subtraction in Multispe...Online Stochastic Tensor Decomposition for Background Subtraction in Multispe...
Online Stochastic Tensor Decomposition for Background Subtraction in Multispe...ActiveEon
 
2010 deep learning and unsupervised feature learning
2010 deep learning and unsupervised feature learning2010 deep learning and unsupervised feature learning
2010 deep learning and unsupervised feature learningVan Thanh
 
NLP Classifier Models & Metrics
NLP Classifier Models & MetricsNLP Classifier Models & Metrics
NLP Classifier Models & MetricsSanghamitra Deb
 
Keynote at IWLS 2017
Keynote at IWLS 2017Keynote at IWLS 2017
Keynote at IWLS 2017Manish Pandey
 
Machine learning and linear regression programming
Machine learning and linear regression programmingMachine learning and linear regression programming
Machine learning and linear regression programmingSoumya Mukherjee
 
An Introduction to Deep Learning
An Introduction to Deep LearningAn Introduction to Deep Learning
An Introduction to Deep Learningmilad abbasi
 
Introduction to Deep Learning
Introduction to Deep LearningIntroduction to Deep Learning
Introduction to Deep LearningMehrnaz Faraz
 
Beyond Low Rank + Sparse: Multi-scale Low Rank Matrix Decomposition
Beyond Low Rank + Sparse: Multi-scale Low Rank Matrix DecompositionBeyond Low Rank + Sparse: Multi-scale Low Rank Matrix Decomposition
Beyond Low Rank + Sparse: Multi-scale Low Rank Matrix DecompositionFrank Ong
 

Similaire à convolutional_rbm.ppt (20)

Deep learning from a novice perspective
Deep learning from a novice perspectiveDeep learning from a novice perspective
Deep learning from a novice perspective
 
Multi-class Classification on Riemannian Manifolds for Video Surveillance
Multi-class Classification on Riemannian Manifolds for Video SurveillanceMulti-class Classification on Riemannian Manifolds for Video Surveillance
Multi-class Classification on Riemannian Manifolds for Video Surveillance
 
CTF: Anomaly Detection in High-Dimensional Time Series with Coarse-to-Fine Mo...
CTF: Anomaly Detection in High-Dimensional Time Series with Coarse-to-Fine Mo...CTF: Anomaly Detection in High-Dimensional Time Series with Coarse-to-Fine Mo...
CTF: Anomaly Detection in High-Dimensional Time Series with Coarse-to-Fine Mo...
 
Comparison of Fine-tuning and Extension Strategies for Deep Convolutional Neu...
Comparison of Fine-tuning and Extension Strategies for Deep Convolutional Neu...Comparison of Fine-tuning and Extension Strategies for Deep Convolutional Neu...
Comparison of Fine-tuning and Extension Strategies for Deep Convolutional Neu...
 
lecture15-neural-nets (2).pptx
lecture15-neural-nets (2).pptxlecture15-neural-nets (2).pptx
lecture15-neural-nets (2).pptx
 
A Re-evaluation of Pedestrian Detection on Riemannian Manifolds
A Re-evaluation of Pedestrian Detection on Riemannian ManifoldsA Re-evaluation of Pedestrian Detection on Riemannian Manifolds
A Re-evaluation of Pedestrian Detection on Riemannian Manifolds
 
IDS for IoT.pptx
IDS for IoT.pptxIDS for IoT.pptx
IDS for IoT.pptx
 
Convolutional Neural Networks (CNN)
Convolutional Neural Networks (CNN)Convolutional Neural Networks (CNN)
Convolutional Neural Networks (CNN)
 
Content-Based Image Retrieval (D2L6 Insight@DCU Machine Learning Workshop 2017)
Content-Based Image Retrieval (D2L6 Insight@DCU Machine Learning Workshop 2017)Content-Based Image Retrieval (D2L6 Insight@DCU Machine Learning Workshop 2017)
Content-Based Image Retrieval (D2L6 Insight@DCU Machine Learning Workshop 2017)
 
Leveraging high level and low-level features for multimedia event detection.2...
Leveraging high level and low-level features for multimedia event detection.2...Leveraging high level and low-level features for multimedia event detection.2...
Leveraging high level and low-level features for multimedia event detection.2...
 
Fundamental of deep learning
Fundamental of deep learningFundamental of deep learning
Fundamental of deep learning
 
General Tips for participating Kaggle Competitions
General Tips for participating Kaggle CompetitionsGeneral Tips for participating Kaggle Competitions
General Tips for participating Kaggle Competitions
 
Online Stochastic Tensor Decomposition for Background Subtraction in Multispe...
Online Stochastic Tensor Decomposition for Background Subtraction in Multispe...Online Stochastic Tensor Decomposition for Background Subtraction in Multispe...
Online Stochastic Tensor Decomposition for Background Subtraction in Multispe...
 
2010 deep learning and unsupervised feature learning
2010 deep learning and unsupervised feature learning2010 deep learning and unsupervised feature learning
2010 deep learning and unsupervised feature learning
 
NLP Classifier Models & Metrics
NLP Classifier Models & MetricsNLP Classifier Models & Metrics
NLP Classifier Models & Metrics
 
Keynote at IWLS 2017
Keynote at IWLS 2017Keynote at IWLS 2017
Keynote at IWLS 2017
 
Machine learning and linear regression programming
Machine learning and linear regression programmingMachine learning and linear regression programming
Machine learning and linear regression programming
 
An Introduction to Deep Learning
An Introduction to Deep LearningAn Introduction to Deep Learning
An Introduction to Deep Learning
 
Introduction to Deep Learning
Introduction to Deep LearningIntroduction to Deep Learning
Introduction to Deep Learning
 
Beyond Low Rank + Sparse: Multi-scale Low Rank Matrix Decomposition
Beyond Low Rank + Sparse: Multi-scale Low Rank Matrix DecompositionBeyond Low Rank + Sparse: Multi-scale Low Rank Matrix Decomposition
Beyond Low Rank + Sparse: Multi-scale Low Rank Matrix Decomposition
 

Dernier

定制(UI学位证)爱达荷大学毕业证成绩单原版一比一
定制(UI学位证)爱达荷大学毕业证成绩单原版一比一定制(UI学位证)爱达荷大学毕业证成绩单原版一比一
定制(UI学位证)爱达荷大学毕业证成绩单原版一比一ss ss
 
1:1原版定制美国加州州立大学东湾分校毕业证成绩单pdf电子版制作修改#真实留信入库#永久存档#真实可查#diploma#degree
1:1原版定制美国加州州立大学东湾分校毕业证成绩单pdf电子版制作修改#真实留信入库#永久存档#真实可查#diploma#degree1:1原版定制美国加州州立大学东湾分校毕业证成绩单pdf电子版制作修改#真实留信入库#永久存档#真实可查#diploma#degree
1:1原版定制美国加州州立大学东湾分校毕业证成绩单pdf电子版制作修改#真实留信入库#永久存档#真实可查#diploma#degreeyuu sss
 
Beautiful Sapna Call Girls CP 9711199012 ☎ Call /Whatsapps
Beautiful Sapna Call Girls CP 9711199012 ☎ Call /WhatsappsBeautiful Sapna Call Girls CP 9711199012 ☎ Call /Whatsapps
Beautiful Sapna Call Girls CP 9711199012 ☎ Call /Whatsappssapnasaifi408
 
Vip Udupi Call Girls 7001305949 WhatsApp Number 24x7 Best Services
Vip Udupi Call Girls 7001305949 WhatsApp Number 24x7 Best ServicesVip Udupi Call Girls 7001305949 WhatsApp Number 24x7 Best Services
Vip Udupi Call Girls 7001305949 WhatsApp Number 24x7 Best Servicesnajka9823
 
Call Girls in Dwarka Sub City 💯Call Us 🔝8264348440🔝
Call Girls in Dwarka Sub City 💯Call Us 🔝8264348440🔝Call Girls in Dwarka Sub City 💯Call Us 🔝8264348440🔝
Call Girls in Dwarka Sub City 💯Call Us 🔝8264348440🔝soniya singh
 
NO1 WorldWide kala jadu Love Marriage Black Magic Punjab Powerful Black Magic...
NO1 WorldWide kala jadu Love Marriage Black Magic Punjab Powerful Black Magic...NO1 WorldWide kala jadu Love Marriage Black Magic Punjab Powerful Black Magic...
NO1 WorldWide kala jadu Love Marriage Black Magic Punjab Powerful Black Magic...Amil baba
 
原版1:1复刻斯坦福大学毕业证Stanford毕业证留信学历认证
原版1:1复刻斯坦福大学毕业证Stanford毕业证留信学历认证原版1:1复刻斯坦福大学毕业证Stanford毕业证留信学历认证
原版1:1复刻斯坦福大学毕业证Stanford毕业证留信学历认证gwhohjj
 
NO1 Qualified Best Black Magic Specialist Near Me Spiritual Healer Powerful L...
NO1 Qualified Best Black Magic Specialist Near Me Spiritual Healer Powerful L...NO1 Qualified Best Black Magic Specialist Near Me Spiritual Healer Powerful L...
NO1 Qualified Best Black Magic Specialist Near Me Spiritual Healer Powerful L...Amil baba
 
专业一比一美国旧金山艺术学院毕业证成绩单pdf电子版制作修改#真实工艺展示#真实防伪#diploma#degree
专业一比一美国旧金山艺术学院毕业证成绩单pdf电子版制作修改#真实工艺展示#真实防伪#diploma#degree专业一比一美国旧金山艺术学院毕业证成绩单pdf电子版制作修改#真实工艺展示#真实防伪#diploma#degree
专业一比一美国旧金山艺术学院毕业证成绩单pdf电子版制作修改#真实工艺展示#真实防伪#diploma#degreeyuu sss
 
《1:1仿制麦克马斯特大学毕业证|订制麦克马斯特大学文凭》
《1:1仿制麦克马斯特大学毕业证|订制麦克马斯特大学文凭》《1:1仿制麦克马斯特大学毕业证|订制麦克马斯特大学文凭》
《1:1仿制麦克马斯特大学毕业证|订制麦克马斯特大学文凭》o8wvnojp
 
Call Girls Delhi {Rohini} 9711199012 high profile service
Call Girls Delhi {Rohini} 9711199012 high profile serviceCall Girls Delhi {Rohini} 9711199012 high profile service
Call Girls Delhi {Rohini} 9711199012 high profile servicerehmti665
 
Gaya Call Girls #9907093804 Contact Number Escorts Service Gaya
Gaya Call Girls #9907093804 Contact Number Escorts Service GayaGaya Call Girls #9907093804 Contact Number Escorts Service Gaya
Gaya Call Girls #9907093804 Contact Number Escorts Service Gayasrsj9000
 
Real Sure (Call Girl) in I.G.I. Airport 8377087607 Hot Call Girls In Delhi NCR
Real Sure (Call Girl) in I.G.I. Airport 8377087607 Hot Call Girls In Delhi NCRReal Sure (Call Girl) in I.G.I. Airport 8377087607 Hot Call Girls In Delhi NCR
Real Sure (Call Girl) in I.G.I. Airport 8377087607 Hot Call Girls In Delhi NCRdollysharma2066
 
Dubai Call Girls O525547819 Spring Break Fast Call Girls Dubai
Dubai Call Girls O525547819 Spring Break Fast Call Girls DubaiDubai Call Girls O525547819 Spring Break Fast Call Girls Dubai
Dubai Call Girls O525547819 Spring Break Fast Call Girls Dubaikojalkojal131
 
Slim Call Girls Service Badshah Nagar * 9548273370 Naughty Call Girls Service...
Slim Call Girls Service Badshah Nagar * 9548273370 Naughty Call Girls Service...Slim Call Girls Service Badshah Nagar * 9548273370 Naughty Call Girls Service...
Slim Call Girls Service Badshah Nagar * 9548273370 Naughty Call Girls Service...nagunakhan
 
Papular No 1 Online Istikhara Amil Baba Pakistan Amil Baba In Karachi Amil B...
Papular No 1 Online Istikhara Amil Baba Pakistan  Amil Baba In Karachi Amil B...Papular No 1 Online Istikhara Amil Baba Pakistan  Amil Baba In Karachi Amil B...
Papular No 1 Online Istikhara Amil Baba Pakistan Amil Baba In Karachi Amil B...Authentic No 1 Amil Baba In Pakistan
 
办理(CSU毕业证书)澳洲查理斯特大学毕业证成绩单原版一比一
办理(CSU毕业证书)澳洲查理斯特大学毕业证成绩单原版一比一办理(CSU毕业证书)澳洲查理斯特大学毕业证成绩单原版一比一
办理(CSU毕业证书)澳洲查理斯特大学毕业证成绩单原版一比一diploma 1
 

Dernier (20)

定制(UI学位证)爱达荷大学毕业证成绩单原版一比一
定制(UI学位证)爱达荷大学毕业证成绩单原版一比一定制(UI学位证)爱达荷大学毕业证成绩单原版一比一
定制(UI学位证)爱达荷大学毕业证成绩单原版一比一
 
young call girls in Khanpur,🔝 9953056974 🔝 escort Service
young call girls in  Khanpur,🔝 9953056974 🔝 escort Serviceyoung call girls in  Khanpur,🔝 9953056974 🔝 escort Service
young call girls in Khanpur,🔝 9953056974 🔝 escort Service
 
1:1原版定制美国加州州立大学东湾分校毕业证成绩单pdf电子版制作修改#真实留信入库#永久存档#真实可查#diploma#degree
1:1原版定制美国加州州立大学东湾分校毕业证成绩单pdf电子版制作修改#真实留信入库#永久存档#真实可查#diploma#degree1:1原版定制美国加州州立大学东湾分校毕业证成绩单pdf电子版制作修改#真实留信入库#永久存档#真实可查#diploma#degree
1:1原版定制美国加州州立大学东湾分校毕业证成绩单pdf电子版制作修改#真实留信入库#永久存档#真实可查#diploma#degree
 
Low rate Call girls in Delhi Justdial | 9953330565
Low rate Call girls in Delhi Justdial | 9953330565Low rate Call girls in Delhi Justdial | 9953330565
Low rate Call girls in Delhi Justdial | 9953330565
 
Beautiful Sapna Call Girls CP 9711199012 ☎ Call /Whatsapps
Beautiful Sapna Call Girls CP 9711199012 ☎ Call /WhatsappsBeautiful Sapna Call Girls CP 9711199012 ☎ Call /Whatsapps
Beautiful Sapna Call Girls CP 9711199012 ☎ Call /Whatsapps
 
Vip Udupi Call Girls 7001305949 WhatsApp Number 24x7 Best Services
Vip Udupi Call Girls 7001305949 WhatsApp Number 24x7 Best ServicesVip Udupi Call Girls 7001305949 WhatsApp Number 24x7 Best Services
Vip Udupi Call Girls 7001305949 WhatsApp Number 24x7 Best Services
 
young call girls in Gtb Nagar,🔝 9953056974 🔝 escort Service
young call girls in Gtb Nagar,🔝 9953056974 🔝 escort Serviceyoung call girls in Gtb Nagar,🔝 9953056974 🔝 escort Service
young call girls in Gtb Nagar,🔝 9953056974 🔝 escort Service
 
Call Girls in Dwarka Sub City 💯Call Us 🔝8264348440🔝
Call Girls in Dwarka Sub City 💯Call Us 🔝8264348440🔝Call Girls in Dwarka Sub City 💯Call Us 🔝8264348440🔝
Call Girls in Dwarka Sub City 💯Call Us 🔝8264348440🔝
 
NO1 WorldWide kala jadu Love Marriage Black Magic Punjab Powerful Black Magic...
NO1 WorldWide kala jadu Love Marriage Black Magic Punjab Powerful Black Magic...NO1 WorldWide kala jadu Love Marriage Black Magic Punjab Powerful Black Magic...
NO1 WorldWide kala jadu Love Marriage Black Magic Punjab Powerful Black Magic...
 
原版1:1复刻斯坦福大学毕业证Stanford毕业证留信学历认证
原版1:1复刻斯坦福大学毕业证Stanford毕业证留信学历认证原版1:1复刻斯坦福大学毕业证Stanford毕业证留信学历认证
原版1:1复刻斯坦福大学毕业证Stanford毕业证留信学历认证
 
NO1 Qualified Best Black Magic Specialist Near Me Spiritual Healer Powerful L...
NO1 Qualified Best Black Magic Specialist Near Me Spiritual Healer Powerful L...NO1 Qualified Best Black Magic Specialist Near Me Spiritual Healer Powerful L...
NO1 Qualified Best Black Magic Specialist Near Me Spiritual Healer Powerful L...
 
专业一比一美国旧金山艺术学院毕业证成绩单pdf电子版制作修改#真实工艺展示#真实防伪#diploma#degree
专业一比一美国旧金山艺术学院毕业证成绩单pdf电子版制作修改#真实工艺展示#真实防伪#diploma#degree专业一比一美国旧金山艺术学院毕业证成绩单pdf电子版制作修改#真实工艺展示#真实防伪#diploma#degree
专业一比一美国旧金山艺术学院毕业证成绩单pdf电子版制作修改#真实工艺展示#真实防伪#diploma#degree
 
《1:1仿制麦克马斯特大学毕业证|订制麦克马斯特大学文凭》
《1:1仿制麦克马斯特大学毕业证|订制麦克马斯特大学文凭》《1:1仿制麦克马斯特大学毕业证|订制麦克马斯特大学文凭》
《1:1仿制麦克马斯特大学毕业证|订制麦克马斯特大学文凭》
 
Call Girls Delhi {Rohini} 9711199012 high profile service
Call Girls Delhi {Rohini} 9711199012 high profile serviceCall Girls Delhi {Rohini} 9711199012 high profile service
Call Girls Delhi {Rohini} 9711199012 high profile service
 
Gaya Call Girls #9907093804 Contact Number Escorts Service Gaya
Gaya Call Girls #9907093804 Contact Number Escorts Service GayaGaya Call Girls #9907093804 Contact Number Escorts Service Gaya
Gaya Call Girls #9907093804 Contact Number Escorts Service Gaya
 
Real Sure (Call Girl) in I.G.I. Airport 8377087607 Hot Call Girls In Delhi NCR
Real Sure (Call Girl) in I.G.I. Airport 8377087607 Hot Call Girls In Delhi NCRReal Sure (Call Girl) in I.G.I. Airport 8377087607 Hot Call Girls In Delhi NCR
Real Sure (Call Girl) in I.G.I. Airport 8377087607 Hot Call Girls In Delhi NCR
 
Dubai Call Girls O525547819 Spring Break Fast Call Girls Dubai
Dubai Call Girls O525547819 Spring Break Fast Call Girls DubaiDubai Call Girls O525547819 Spring Break Fast Call Girls Dubai
Dubai Call Girls O525547819 Spring Break Fast Call Girls Dubai
 
Slim Call Girls Service Badshah Nagar * 9548273370 Naughty Call Girls Service...
Slim Call Girls Service Badshah Nagar * 9548273370 Naughty Call Girls Service...Slim Call Girls Service Badshah Nagar * 9548273370 Naughty Call Girls Service...
Slim Call Girls Service Badshah Nagar * 9548273370 Naughty Call Girls Service...
 
Papular No 1 Online Istikhara Amil Baba Pakistan Amil Baba In Karachi Amil B...
Papular No 1 Online Istikhara Amil Baba Pakistan  Amil Baba In Karachi Amil B...Papular No 1 Online Istikhara Amil Baba Pakistan  Amil Baba In Karachi Amil B...
Papular No 1 Online Istikhara Amil Baba Pakistan Amil Baba In Karachi Amil B...
 
办理(CSU毕业证书)澳洲查理斯特大学毕业证成绩单原版一比一
办理(CSU毕业证书)澳洲查理斯特大学毕业证成绩单原版一比一办理(CSU毕业证书)澳洲查理斯特大学毕业证成绩单原版一比一
办理(CSU毕业证书)澳洲查理斯特大学毕业证成绩单原版一比一
 

convolutional_rbm.ppt

  • 1. Convolutional Restricted Boltzmann Machines for Feature Learning Mohammad Norouzi Advisor: Dr. Greg Mori CS @ Simon Fraser University 27 Nov 2009 1
  • 2. CRBMs for Feature Learning Mohammad Norouzi Advisor: Dr. Greg Mori CS @ Simon Fraser University 27 Nov 2009 2
  • 5. Sliding Window Approach (Cont’d) 5 [INRIA Person Dataset]
  • 6. Success or Failure of an object recognition algorithm hinges on the features used Input Feature representation Label Our Focus Classifier ? Human Background 0 / 1 / 2 / 3 / … 6 Learning
  • 7. Local Feature Detector Hierarchies 7 Larger More complicated Less frequent
  • 8. Generative & Layerwise Learning 8 ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? Generative CRBM ? ? ? ? ? ? ? ? ? ? ? ?
  • 9. Visual Features: Filtering 9 1 0 -1 2 0 -2 1 0 -1 Filter Kernel (Feature) -1 0 1 -2 0 2 -1 0 1 0 -1 -2 1 0 -1 2 1 0 Filter Response 1 W V 2 W 2 W ) , ( 1 W V Filter ) , ( 2 W V Filter ) , ( 3 W V Filter
  • 10. Our approach to feature learning is generative ? ? ? 1 H 2 H 3 H V Binary Hidden Variables 10 1 W 2 W 3 W (CRBM model)
  • 12. Related Work • Convolutional Neural Network (CNN) – Filtering layers are bundled with a classifier, and all the layers are learned together using error backpropagation. – Does not perform well on natural images • Biologically plausible models – Hand-crafted first layer vs. Randomly selected prototypes for second layer. [Lecun et al. 98] [Ranzato et al. CVPR'07] [Serre et al., PAMI'07] [Mutch and Lowe, CVPR'06] 12
  • 13. Related Work (cont’d) • Deep Belief Net – A two layer partially observed MRF, called RBM, is the building block – Learning is performed unsupervised and layer-by- layer from bottom layer upwards • Our contributions: We incorporate spatial locality into RBMs and adapt the learning algorithm accordingly • We add more complicated components such as pooling and sparsity into deep belief nets [Hinton et al., NC'2006] 13
  • 14. Why Generative &Unsupervised • Discriminative learning of deep and large neural networks has not been successful – Requires large training sets – Easily gets over-fitted for large models – First layer gradients are relatively small • Alternative hybrid approach – Learn a large set of first layer features generatively – Switch to a discriminative model to select the discriminative features from those that are learned – Discriminative fine-tuning is helpful
  • 16. CRBM • Image is the visible layer and hidden layer is related to filter responses • An energy based probabilistic model 16 Dot product of vectorized matrices       ) , ( ) ; , ( ) ; , ( ) ; , ( , exp 1 k k k k k k k H W V Filter H W H V E W H V E W H V E H;W V E Z = V;W P       
  • 17. Training CRBMs • Maximum likelihood learning of CRBMs is difficult • Contrastive Divergence (CD) learning is applicable • For CD learning we need to compute the conditionals and . data 17 sample   H V P |   V H P |
  • 18. CRBM (Backward) • Nearby hidden variables cooperate in reconstruction • Conditional Probabilities take the form 18     ) exp 1 ( 1 * ) ( ) , ( ) | ( ) , ( ) | ( x k k k k k x W H Filter H V P W V Filter V H P         
  • 19. Learning the Hierarchy • The structure is trained bottom up and layerwise • The CRBM model for training filtering layers • Filtering layers are followed by down-sampling CRBM CRBM Classifier Pooling Pooling 19 Filtering Non-linearity Reduce the dimensionality layers
  • 22. Evaluation MNIST digit dataset • Training set: 60,000 image of digits of size 28x28 • Test set: 10,000 images INRIA person dataset • Training set: 2416 person windows of size 128 x 64 pixels and 4.5x106 negative windows • Test set: 1132 positive and 2x106 negative windows 22
  • 23. First layer filters • Gray-scale images of INRIA positive set • 15 filters of 7x7 23 • MNIST unlabeled digits • 15 filters of 5x5
  • 24. Second Layer Features (MNIST) • Hard to visualize the filters • We show patches highly responded to filters: 24 24
  • 25. Second Layer Features (INRIA) 25
  • 26. MNIST Results • MNIST error rate when model is trained on the full training set 26
  • 33. INRIA Results • Adding our large-scale features significantly improves performance of the baseline (HOG) 33
  • 34. Conclusion • We extended the RBM model to Convolutional RBM, useful for domains with spatial locality • We exploited CRBMs to train local hierarchical feature detectors one layer at a time and generatively • This method obtained results comparable to state-of-the-art in digit classification and human detection 34
  • 37. Contrastive Divergence Learning 37   data 1 k data 0 k k k H , V Filter H , V Filter η + W = W ) ( ) ( 1 0      k k H V, Filter = W θ H; V, E   
  • 38. Training CRBMs (Cont'd) • The problem of reconstructing border region becomes severe when number of Gibbs sampling steps > 1. – Partition visible units into middle and border regions • Instead of maximizing the likelihood, we (approximately) maximize    b m v | v p
  • 39. Enforcing Feature Sparsity • The CRBM's representation is K (number of filters) times overcomplete • After a few CD learning iterations, V is perfectly reconstructed • Enforce sparsity to tackle this problem – Hidden bias terms were frozen at large negative values • Having a single non-sparse hidden unit improves the learned features – Might be related to the ergodicity condition
  • 40. Probabilistic Meaning of Max 1 2 3 4 5 6 1 2 3 4 Max 1 2 3 4 5 6 1 1 2 2 h h' v   6 4 5 3 4 2 3 1 : T 4 : T 3 : T 2 : T 1 v w h + v w h + v w h + v w h = h v, E  h' v       6 4 5 3 4 2 3 1 : T 2 : T 2 : T 1 : T 1 v w h' + v w h' max + v w h' , v w h' max = h v, E 
  • 41. The Classifier Layer • We used SVM as our final classifier – RBF kernel for MNIST – Linear kernel for INRIA – For INRIA we combined our 4th layer outputs and HOG features • We experimentally observed that relaxing the sparsity of CRBM's hidden units yields better results – This lets the discriminative model to set the thresholds itself
  • 42. Why HOG features are added? • Because part-like features are very sparse • Having a template of the human figure helps a lot f
  • 43. RBM • Two layer pairwise MRF with a full set of hidden-visible connections • RBM Is an energy based model • Hidden random variables are binary, Visible variables can be binary or continuous • Inference is straightforward: and • Contrastive Divergence learning for training h v w         θ h; v, E θ Z = θ h; v, p  exp 1          2 2 1 i j j i i j ij i v + h c v b h w v = θ h; v, E   v | h p   h | v p
  • 44. Why Unsupervised Bottom-Up • Discriminative learning of deep structure has not been successful – Requires large training sets – Easily is over-fitted for large models – First layer gradients are relatively small • Alternative hybrid approach – Learn a large set of first layer features generatively – Later, switch to a discriminative model to select the discriminative features from those learned – Fine-tune the features using
  • 45. INRIA Results (Cont'd) • Missrate at different FPPW rates • FPPI is a better indicator of performance • More experiments on size of features and number of layers are desired