The document outlines seven essential steps for creating effective deep neural networks, beginning with the single layer perceptron and moving through multi-layer perceptrons, convolutional networks, and regularization techniques. It emphasizes the importance of batch normalization, inception architectures, and residual networks, culminating in the use of LSTM for image processing. Additionally, it provides various references and resources for further learning in neural network design and implementation.

1. 7 steps for highly effective
deep neural networks
Natalino Busa - Head of Data Science

2. 2 Natalino Busa - @natbusa
O’Reilly Author and Speaker
Teradata EMEA Practice Lead on Open Source Technologies
Teradata Principal Data Scientist
ING Group Enterprise Architect: Cybersecurity, Marketing, Fintech
Cognitive Finance Group Advisory Board Member
Philips Senior Researcher, Data Architect
Linkedin and Twitter:
@natbusa

3. 3
Natalino Busa - @natbusa
X: independent variable
Y: dependent variable
Linear Regression:
How to best fit a line to some data

4. 4
Natalino Busa - @natbusa
input : 784 numbers
output: 10 classes
28x28 pixels
Classification: Handwritten digits

5. 5
Natalino Busa - @natbusa
Sharing the (Not so) Secret Lore:
Keras & Tensorflow
Some smaller projects:
Tflearn, Tensorlayers
http://keras.io/

6. 6
Natalino Busa - @natbusa
1: Single Layer Perceptron
“dendrites”
Axon’s
response
Activation function

7. 7
Natalino Busa - @natbusa
More activation functions:
https://en.wikipedia.org/wiki/Activation_function

8. 8
Natalino Busa - @natbusa
Single Layer Neural Network
Takes: n-input features: Map them to a soft “binary” space
∑
x1
x2
xn
f
1: Single Layer Perceptron (binary classifier)

9. 9
Natalino Busa - @natbusa
1: Single Layer Perceptron (multi-class classifier)
Values between 0 and 1
Sum of all outcomes = 1
It produces an estimate of a probability!
From soft binary space to predicting probabilities:
Take n inputs, Divide by the sum of the predicted values
∑
x1
x2
xn
f
∑ f
softmax ‘1’: 95%
‘8’: 5%

10. 10
Natalino Busa - @natbusa
1: Single Layer Perceptron
Minimize costs:
The cost function depends on:
- Parameters of the model
- How the model “composes”
Goal :
Reduce the mean probability error
modify the parameters to reduce the error! Vintage math from last century

11. 11
Natalino Busa - @natbusa
Supervised Learning
Stack layers of perceptrons
- Feed Forward Network
- Scoring goes
from input to output
- Back propagate the error
from output to input
SOFTMAX
Input parameters
classes (estimated probabilities)
Feed-forwardfunctions
Cost function
supervised : actual output
BackPropagateErrors

12. 12
Natalino Busa - @natbusa
Let’s go!

13. 13
Natalino Busa - @natbusa
1. Single Layer Perceptron

14. 14
Natalino Busa - @natbusa
1. Single Layer Perceptron

15. 15
Natalino Busa - @natbusa
1. Single Layer Perceptron

16. 16
Natalino Busa - @natbusa
1. Single Layer Perceptron

17. 17
Natalino Busa - @natbusa
Tensorboard

18. 18
Natalino Busa - @natbusa
Tensorboard

19. 19
Natalino Busa - @natbusa
Tensorboard

20. 20
Natalino Busa - @natbusa
2. Multi Layer Perceptron

21. 21
Natalino Busa - @natbusa
2. Multi Layer Perceptron

22. 22
Natalino Busa - @natbusa
2. Multi Layer Perceptron

23. 23
Natalino Busa - @natbusa
2. Multi Layer Perceptron

24. 24
Natalino Busa - @natbusa
3. Convolution
From Krizehvsky et al. (2012)

25. 25
Natalino Busa - @natbusa
3. Convolution
diagrams:
By Aphex34 - Own work, CC BY-SA 4.0, https://commons.wikimedia.org/w/index.php?curid=45659236
By Aphex34 - Own work, CC BY-SA 4.0, https://commons.wikimedia.org/w/index.php?curid=45673581
CC0, https://en.wikipedia.org/w/index.php?curid=48817276
convolution Max pooling RELU / ELU

26. 26
Natalino Busa - @natbusa
3. Convolution

27. 27
Natalino Busa - @natbusa
3. Convolution

28. 28
Natalino Busa - @natbusa
3. Batch Normalization

29. 29
Natalino Busa - @natbusa
3. Batch Normalization ( example for MLP)

30. 30
Natalino Busa - @natbusa
3. Batch Normalization ( example for MLP)

31. 31
Natalino Busa - @natbusa
3. Batch Normalization ( example for MLP)
Activation function:
SIGMOID

32. 32
Natalino Busa - @natbusa
3. Batch Normalization ( example for MLP)
Activation function:
RELU

33. 33
Natalino Busa - @natbusa
4. Regularization: Prevent overfitting in ANNs
- Batch Normalization
- RELU/ELU
- RESIDUAL / SKIP Networks
- DROP LAYER
- REDUCE PRECISION (HUFFMAN ENCODING)
In general ANN are parameters rich, constraining the
parameter space usually produces better results and speed
up the learning

34. 34
Natalino Busa - @natbusa
5. Inception architectures
Cannot be stacked!

35. 35
Natalino Busa - @natbusa
5. Inception architectures

36. 36
Natalino Busa - @natbusa
5. Inception architectures
compression
Avoid
dimensions
explosion

37. 37
Natalino Busa - @natbusa
5. Inception architectures

38. 38
Natalino Busa - @natbusa
5. Inception architectures (top architecture)

39. 39
Natalino Busa - @natbusa
6. Residual Networks
https://culurciello.github.io/tech/2016/06/04/nets.html

40. 40
Natalino Busa - @natbusa
6. Residual Networks

41. 41
Natalino Busa - @natbusa
6. Residual + Inception Networks

42. 42
Natalino Busa - @natbusa
7. LSTM on Images

43. 43
Natalino Busa - @natbusa
7. LSTM on Images

44. 44
Natalino Busa - @natbusa
7. LSTM on Images
http://colah.github.io/posts/2015-08-Understanding-LSTMs/

45. 45
Natalino Busa - @natbusa
7. LSTM on Images

46. 46
Natalino Busa - @natbusa
7. LSTM on Images

47. 47
Natalino Busa - @natbusa
7. LSTM on ConvNets (bonus slide)

48. 48
Natalino Busa - @natbusa
All the codez :)
https://github.com/natbusa/deepnumbers

49. 49
Natalino Busa - @natbusa
Meta- References
… just a few articles, but extremely dense in content.
A must read!
https://keras.io/
http://karpathy.github.io/neuralnets/
https://culurciello.github.io/tech/2016/06/04/nets.html
http://colah.github.io/posts/2015-08-Understanding-LSTMs/
https://gab41.lab41.org/batch-normalization-what-the-hey-d480039a9e3b