VIP Call Girls Service Hitech City Hyderabad Call +91-8250192130
Introduction to Deep Learning with TensorFlow
1. Terry Taewoong Um (terry.t.um@gmail.com)
University of Waterloo
Department of Electrical & Computer Engineering
Terry Taewoong Um
INTRODUCTION TO DEEP
NEURAL NETWORK WITH
TENSORFLOW
1
2. Terry Taewoong Um (terry.t.um@gmail.com)
CONTENTS
2
1. Why Deep Neural Network
3. Terry Taewoong Um (terry.t.um@gmail.com)
3
EXAMPLE CASE
- Imagine you have extracted features from sensors
- The dimension of each sample (which represents
one of gestures) is around 800
- You have 70,000 samples (trial)
- What method would you apply?
4. Terry Taewoong Um (terry.t.um@gmail.com)
4
EXAMPLE CASE
- Reduce the dimension from 800 to 40 by using a
feature selection or dim. reduction technique
☞ What you did here is “Finding a good representation”
- Then, you may apply a classification methods to
classify 10 classes
• You may have several ways to do it
• But, what if
- You have no idea for feature selection?
- The dimension is much higher than 800 and
you have more classes.
?
5. Terry Taewoong Um (terry.t.um@gmail.com)
5
EXAMPLE CASE
- Reduce the dimension from 800 to 40 by using a
feature selection or dim. reduction technique
☞ What you did here is “Finding a good representation”
- Then, you may apply a classification methods to
classify 10 classes
• You may have several ways to do it
• But, what if
- You have no idea for feature selection?
- The dimension is much higher than 800 and
you have more classes.
MNIST dataset
(65000spls * 784dim)
MNIST dataset
(60000spls * 1024dim)
6. Terry Taewoong Um (terry.t.um@gmail.com)
6
CLASSIFICATION RESULTS
error rate : 28% → 15% → 8%
(2010) (2014)(2012)
http://rodrigob.github.io/are_we_there_yet/bu
ild/classification_datasets_results.html
7. Terry Taewoong Um (terry.t.um@gmail.com)
7
PARADIGM CHANGE
Knowledge
PRESENT
Representation
(Features)
How can we find a
good representation?
IMAGE
SPEECH
Hand-Crafted Features
8. Terry Taewoong Um (terry.t.um@gmail.com)
8
PARADIGM CHANGE
IMAGE
SPEECH
Hand-Crafted Features
Knowledge
PRESENT
Representation
(Features)
Can we learn a good representation
(feature) for the target task as well?
9. Terry Taewoong Um (terry.t.um@gmail.com)
9
UNSUPERVISED LEARNING
“Convolutional deep belief networks for scalable unsupervised learning of hierarchical representation”, Lee et al., 2012
10. Terry Taewoong Um (terry.t.um@gmail.com)
10
THREE TYPES OF DEEP LEARNING
• Unsupervised learning method
Autoencoder http://goo.gl/s6kmqY
- Restricted Boltzmann Machine(RBM), Autoencoder, etc.
- It helps to avoid local minima problem
(It regularizes the training data)
- But it is not necessary when we have large amount of data.
(Drop-out is enough for regularization)
• Convolutional Neural Network (ConvNet)
• Recurrent Neural Network (RNN) + Long-Short Term Memory (LSTM)
- ConvNet has shown outstanding performance in recognition tasks (image, speech)
- ConvNet contains hierarchical abstraction process called pooling.
- RNN+LSTM makes use of long-term memory → Good for time-series data
- RNN is a generative model: It can generate new data
11. Terry Taewoong Um (terry.t.um@gmail.com)
CONTENTS
11
2. DNN with TensorFlow
Thanks to Sungjoon Choi
https://github.com/sjchoi86/
14. Terry Taewoong Um (terry.t.um@gmail.com)
14
DEEP LEARNING LIBRARY
• Karpathy’s Recommendation
15. Terry Taewoong Um (terry.t.um@gmail.com)
15
BASIC WORKFLOW OF TF
1. Load data
2. Define the NN structure
3. Set optimization parameters
4. Run!
https://github.com/terryum/TensorFlow_Exercises
16. Terry Taewoong Um (terry.t.um@gmail.com)
16
EXAMPLE 1
https://github.com/terryum/TensorFlow_Exercises
17. Terry Taewoong Um (terry.t.um@gmail.com)
17
1. LOAD DATA https://github.com/terryum/TensorFlow_Exercises/blob/
master/2_LogisticRegression_MNIST_160516.ipynb
22. Terry Taewoong Um (terry.t.um@gmail.com)
22
EXAMPLE 2
https://github.com/terryum/TensorFlow_Exercises
23. Terry Taewoong Um (terry.t.um@gmail.com)
23
NEURAL NETWORK
Hugo Larochelle, http://www.dmi.usherb.ca/~larocheh/index_en.html
• Activation functions
http://goo.gl/qMQk5H
• Basic NN structure
24. Terry Taewoong Um (terry.t.um@gmail.com)
24
1. LOAD DATA https://github.com/terryum/TensorFlow_Exercises/blob/
master/3a_MLP_MNIST_160516.ipynb
25. Terry Taewoong Um (terry.t.um@gmail.com)
25
2. DEFINE THE NN STRUCTURE
26. Terry Taewoong Um (terry.t.um@gmail.com)
26
3. SET OPTIMIZATION PARAMETERS
28. Terry Taewoong Um (terry.t.um@gmail.com)
28
EXAMPLE 3
https://github.com/terryum/TensorFlow_Exercises
29. Terry Taewoong Um (terry.t.um@gmail.com)
29
CONVOLUTION
http://colah.github.io/posts/2014-07-
Understanding-Convolutions/
http://www.wildml.com/2015/11/understanding-convolutional-neural-networks-for-nlp/
30. Terry Taewoong Um (terry.t.um@gmail.com)
30
CONVOLUTIONAL NN
• How can we deal with real images which is
much bigger than MNIST digit images?
- Use not fully-connected, but locally-connected NN
- Use convolutions to get various feature maps
- Abstract the results into higher layer by using pooling
- Fine tune with fully-connected NN
https://goo.gl/G7kBjI
https://goo.gl/Xswsbd
http://goo.gl/5OR5oH
31. Terry Taewoong Um (terry.t.um@gmail.com)
31
1. LOAD DATA https://github.com/terryum/TensorFlow_Exercises/blob/
master/4a_CNN_MNIST_160517.ipynb
32. Terry Taewoong Um (terry.t.um@gmail.com)
32
2. DEFINE THE NN STRUCTURE
33. Terry Taewoong Um (terry.t.um@gmail.com)
33
2. DEFINE THE NN STRUCTURE
34. Terry Taewoong Um (terry.t.um@gmail.com)
34
3. SET OPTIMIZATION PARAMETERS