Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.
First step Deep Learning
with Tensorflow
1
Jiqiong QIU
About me
2
● Master of engineer degree at INSA de Lyon
● PhD in Color Formulation by Statistical Learning at UTC
and BASF ...
1. What is Deep Learning?
2. Why Tensorflow?
3. Digital Recognition using CNN
4. Resources
First step Deep Learning with T...
What is Deep Learning?1
4
1 What is Deep Learning
1.1 From AI to Deep Learning
1.2 Neural Network
1.3 Deep Learning
5
1.1 From AI to Deep Learning
Artificial
Intelligence
Machine
Learning
Logistic,
Regression, SVM,
Neural Network
Deep Learn...
1.2 Neural Network
Artificial Neuron
Neuron
7
1.2 Neural Network
8
MLP: Multilayer Perceptron
Hidden layer
Output layer
Input layer
Input #1
Input #3
Input #4
Input #5
...
1.3 Deep Learning
9
● More layer
● Different Neuron types
Output layer
ALEXNET
Why Tensorflow?2
10
2 Why Tensorflow?
11
● Open source by Google
● Python API
● Board
● Android (SDK) Mobile application
And more: http://deep...
Digital Recognition
using CNN3
12
3 Digital Recognition using CNN
3.1 Digital Recognition and Data Set
3.2 CNN
13
3.1 Digital Recognition and Data Set
14
● Images 28x28 pixels
● Training set: 60 000
● Test set: 10 000
MNIST database: Mi...
3.2 CNN : Convolutionnal Neural Network
15
3.2.1 Convolution layer
3.2.2 Pooling layer
3.2.3 MLP and more
3.2.4 Training a...
3.2 Digital Recognition using CNN
16
Convolutionnal Neural Network
3.2.1 Convolution layer
17
Natural Images have the property of being “stationary”
3.2.1 Convolution layer
18
Kernel Matrix
3.2.1 Convolution layer
19
3D Neuron MATRIX
Input Depth = 3 (rgb)
Output Depth = 400
3.2.1 Convolution layer
20
3.2.1 Convolution layer: Tensorflow
21
3.2.2 Pooling layer
22
Pooling (or Subsampling)
● 28x28 pixels
● 400 features (depth) over 3x3 inputs
● 2 strides
● 13x13x...
3.2.2 Pooling layer: Tensorflow
23
3.2.3 MLP and More
24
MLPConvolution + Pooling
3.2.3 MLP and More
25
MLPConvolution + Pooling
Reshape (3D to 1D)
3.2.3 MLP and More
26
Dropout
Hidden layer
Output layer
Input layer
Input #1
Input #3
Input #4
Input #5
Output
3.2.3 MLP and More
27
● Multiple Classification:
[0,1,..,8,9]
● Softmax:
3.2.3 MLP and More: Tensorflow
28
● Reshape
● Dropout
● Output of CNN
3.2.4 Training and Results
29
● Cost Function:
● Cost Function is defined by:
○ Input Values (images, noted as x)
○ Weight...
3.2.4 Training and Results
30
Training process
3.2.4 Training and Results: Tensorflow
31
● Cost Function
● Training process
3.2.4 Training and Results
32
3.2.4 Training and Results
33
3.2.4 Training and Results
34
Resources4
35
4 Resources
● For this tutorial:
https://github.com/Sfeir/demo-tenserflow
https://docs.google.com/presentation/d/1ch4YiKD8...
Thank you.
37
38
by Jiqiong QIU
SFEIR - Copyright ©2016
Upcoming SlideShare
Loading in …5
×

First Step Deep Learning by Jiqiong Qiu - DevFest 2016

3,708 views

Published on

Find Jiquiong Qiu's slides about Deep Learning with Tensorflow at the DevFest Paris 2016

Published in: Software
  • Be the first to comment

First Step Deep Learning by Jiqiong Qiu - DevFest 2016

  1. 1. First step Deep Learning with Tensorflow 1 Jiqiong QIU
  2. 2. About me 2 ● Master of engineer degree at INSA de Lyon ● PhD in Color Formulation by Statistical Learning at UTC and BASF Coatings (EP 2887275A1: Method and system for determining a color formula) ● Data scientist at Sfeir
  3. 3. 1. What is Deep Learning? 2. Why Tensorflow? 3. Digital Recognition using CNN 4. Resources First step Deep Learning with Tensorflow 3
  4. 4. What is Deep Learning?1 4
  5. 5. 1 What is Deep Learning 1.1 From AI to Deep Learning 1.2 Neural Network 1.3 Deep Learning 5
  6. 6. 1.1 From AI to Deep Learning Artificial Intelligence Machine Learning Logistic, Regression, SVM, Neural Network Deep Learning CNN, LSTM, Neural Turing Macines 6
  7. 7. 1.2 Neural Network Artificial Neuron Neuron 7
  8. 8. 1.2 Neural Network 8 MLP: Multilayer Perceptron Hidden layer Output layer Input layer Input #1 Input #3 Input #4 Input #5 Output
  9. 9. 1.3 Deep Learning 9 ● More layer ● Different Neuron types Output layer ALEXNET
  10. 10. Why Tensorflow?2 10
  11. 11. 2 Why Tensorflow? 11 ● Open source by Google ● Python API ● Board ● Android (SDK) Mobile application And more: http://deeplearning.net/software_links/
  12. 12. Digital Recognition using CNN3 12
  13. 13. 3 Digital Recognition using CNN 3.1 Digital Recognition and Data Set 3.2 CNN 13
  14. 14. 3.1 Digital Recognition and Data Set 14 ● Images 28x28 pixels ● Training set: 60 000 ● Test set: 10 000 MNIST database: Mixed National Institute of Standards and Technology database
  15. 15. 3.2 CNN : Convolutionnal Neural Network 15 3.2.1 Convolution layer 3.2.2 Pooling layer 3.2.3 MLP and more 3.2.4 Training and Results
  16. 16. 3.2 Digital Recognition using CNN 16 Convolutionnal Neural Network
  17. 17. 3.2.1 Convolution layer 17 Natural Images have the property of being “stationary”
  18. 18. 3.2.1 Convolution layer 18 Kernel Matrix
  19. 19. 3.2.1 Convolution layer 19 3D Neuron MATRIX Input Depth = 3 (rgb) Output Depth = 400
  20. 20. 3.2.1 Convolution layer 20
  21. 21. 3.2.1 Convolution layer: Tensorflow 21
  22. 22. 3.2.2 Pooling layer 22 Pooling (or Subsampling) ● 28x28 pixels ● 400 features (depth) over 3x3 inputs ● 2 strides ● 13x13x400 = 169x400= 67600
  23. 23. 3.2.2 Pooling layer: Tensorflow 23
  24. 24. 3.2.3 MLP and More 24 MLPConvolution + Pooling
  25. 25. 3.2.3 MLP and More 25 MLPConvolution + Pooling Reshape (3D to 1D)
  26. 26. 3.2.3 MLP and More 26 Dropout Hidden layer Output layer Input layer Input #1 Input #3 Input #4 Input #5 Output
  27. 27. 3.2.3 MLP and More 27 ● Multiple Classification: [0,1,..,8,9] ● Softmax:
  28. 28. 3.2.3 MLP and More: Tensorflow 28 ● Reshape ● Dropout ● Output of CNN
  29. 29. 3.2.4 Training and Results 29 ● Cost Function: ● Cost Function is defined by: ○ Input Values (images, noted as x) ○ Weight matrix (w)
  30. 30. 3.2.4 Training and Results 30 Training process
  31. 31. 3.2.4 Training and Results: Tensorflow 31 ● Cost Function ● Training process
  32. 32. 3.2.4 Training and Results 32
  33. 33. 3.2.4 Training and Results 33
  34. 34. 3.2.4 Training and Results 34
  35. 35. Resources4 35
  36. 36. 4 Resources ● For this tutorial: https://github.com/Sfeir/demo-tenserflow https://docs.google.com/presentation/d/1ch4YiKD83wERmmEFRvFIQ98Mtz65aGsLP3uWiTElY2I/edit? usp=sharing ● MOOC: CS229: Machine Learning UD730: Deep Learning - Taking machine learning to the next level CS231n: Convolutional Neural Networks for Visual Recognition CS224d:Deep Learning for Natural Language Processing ● Summer school: IPAM 2012 Montreal 2015 ● Website http://deeplearning.net/ https://www.topcoder.com/ https://www.kaggle.com/ 36
  37. 37. Thank you. 37
  38. 38. 38 by Jiqiong QIU SFEIR - Copyright ©2016

×