learning
play

Learning Prof. Kuan-Ting Lai 2019/7/2 Deep Learning a new Buzzword - PowerPoint PPT Presentation

Introduction to Deep Learning Prof. Kuan-Ting Lai 2019/7/2 Deep Learning a new Buzzword 2 AI Papers 3 Registration of NIPS 4 AL/ML Investement 5 Source: Sand Hill Econometrics 6 Source: Sand Hill Econometrics 7 AlphaGo 8 So,


  1. Introduction to Deep Learning Prof. Kuan-Ting Lai 2019/7/2

  2. Deep Learning – a new Buzzword 2

  3. AI Papers 3

  4. Registration of NIPS 4

  5. AL/ML Investement 5

  6. Source: Sand Hill Econometrics 6 Source: Sand Hill Econometrics

  7. 7

  8. AlphaGo 8

  9. So, what is Deep Learning?

  10. 10

  11. Machine Learning

  12. 12

  13. 13

  14. Learning Representation • Objective: Classify white & black • Input: (x, y) • Output: Black or White 14

  15. The Master Algorithm – Pedro Domingos 15

  16. Five Tribes of Machine Learning • Evolutionaries ( 基因演化法 ) • Connectionists ( 類神經網路 ) • Symbolists ( 歸納法 ) • Bayesians ( 貝氏機率 ) • Analogizers ( 類比近似 ) 16

  17. Five Tribes of Machine Learning • Symbolists: Decision Trees, Random Forest • Bayesians: Naïve Bayesians • Analogizers: SVM, k-NN • Evolutionaries: Gene algorithms • Connectionists: Deep Learning 17

  18. All Algorithms can be Reduced to 3 Operations 1 1 0 1 1 0 0 0 18

  19. 1 XOR 0 1 19

  20. OK, machine learning is cool. But what is Deep Learning?

  21. 21

  22. Neuron 22

  23. Frank Rosenblatt’s Perceptron (1957) 23

  24. 24

  25. 25

  26. 26

  27. Deep Learning 27

  28. 28

  29. Learning XOR (1986) Geoffrey Hinton 29

  30. Backpropagation 30

  31. Chain Rule 31

  32. Computation Graph c = a + b d = b + 1 e = c*d 32

  33. MNIST database of Handwritten Digits 33

  34. 34

  35. 35

  36. 36

  37. 37

  38. 38

  39. 39

  40. 40

  41. Gradient Descent 41

  42. https://hackernoon.com/gradient-descent-aynk-7cbe95a778da 42

  43. Cost Function • Mean-Squared Error 𝑂 𝐾 𝜄 = 1 𝜄 𝑦 𝑗 − 𝑧 𝑗 2 𝑂 ෍ 𝑔 𝑗=1 43

  44. Gradient Descent of MSE • Gradient of MSE 𝑂 𝜖𝐾 𝜄 = 2 ′ 𝑦 𝑗 𝑂 ෍ 𝑔 𝜄 𝑦 𝑗 − 𝑧 𝑗 𝑔 𝜄 𝜖𝜄 𝑗=1 • Update 𝑘 − 𝛽 𝜖𝐾 𝜄 𝜄 𝑘 ← 𝜄 𝜖𝜄 𝑘 • Repeat until Convergence 44

  45. 45

  46. 46

  47. Convolutional Neural Network (LeNet-5) • https://medium.com/@sh.tsang/paper-brief-review-of-lenet-1-lenet-4-lenet-5- boosted-lenet-4-image-classification-1f5f809dbf17

  48. 48

  49. ImageNet Large Scale Visual Object Recognition Challenge (ILSVRC) • 1000 categories • For ILSVRC 2017 − Training images for each category ranges from 732 to 1300 − 50,000 validation images and 100,000 test images . • Total number of images in ILSVRC 2017 is around 1,150,000 49

  50. Convolutional Neural Network • Alex Krizhevsky, Geoffrey Hinton et al., 2012 50

  51. Previous Winners of ILSVRC 51

  52. Deep Reinforcement Learnin g 52

  53. Reinforcement Learning

  54. 54

  55. AlphaGo 55

  56. The Complexity of Go vs Chess 56

  57. Reinforcement Learning • An agent learns how to do actions a t to achieve maximum reward R • Policy π( a t |s t ) : agent’s behavior function • Value function V : evaluate quality of each action/state • Model: agent’s representation of the environment Policy 57

  58. Learn to Play Atari Games • Mnih et al., “Human Level Control through Deep Reinforcement Learning,” Nature , 2015 58

  59. DRL in Atari

  60. AlphaGo Zero 60

  61. 61

  62. 62

  63. Virtual-to-real Learning • Inspired by DeepMind (Mnih et al., Nature , 2015) − “Human Level Control through Deep Reinforcement Learning” • Applied to computer vision applications − Image segmentation : Armeni et al. (2016), Qiu et al., (2017) − Indoor navigation : Brodeur et al. (2017), Gupta et al. (2017), Savva et al. (2017), Wu et al. (2018) − Autonomous vehicles : Marinez et al. (2017), Muller et al. (2018), Pan et al. (2017), Shah et al. (2018) CAD 2 Real UnrealCV 63

  64. Semantic Segmentation Autonomous Navigation VIVID Depth Prediction Action Recognition 64

  65. Simulate Real-life Events 65

  66. Searching for the Shooter 66

  67. DeepDrive 67

  68. Limits of Deep Learning 68

  69. No Idea of Real World 69

  70. Adversarial Attack 70

  71. Number of Connections in the Brain Neurons (for adults): 10 ^11 , or 100 billion, 100000000000 Synapses (based on 1000 per neuron): 10^ 14 , or 100 trillion, 100000000000000 71

  72. Generative Adversarial Networks (GAN) 72

  73. Generative Adversarial Networks (GAN) • Ian Goodfellow 73

  74. Painting like Van Gogh 74

  75. Super Resolution 75

  76. DeepFake: Is this you? 76

  77. Google’s AutoML • Learning neural network cells automatically https://ai.googleblog.com/2017/11/automl-for-large-scale-image.html 77

  78. AutoML on ImageNet 78

  79. EfficientNet (May, 2019) 79

  80. 80

  81. References • Francois Chollet, “Deep Learning with Python.” Chapter 1 • What is backpropagation really doing? ( 3Blue1Brown) https://www.youtube.com/watch?v=Ilg3gGewQ5U • http://www.andreykurenkov.com/writing/ai/a-brief-history-of-neural-nets-and- deep-learning/ • https://pmirla.github.io/2016/08/16/AI-Winter.html • https://tw.saowen.com/a/6cdc2f1279016e566832bb1234e06d321992dd1fabcdf 4a2e0a3e16fc0dc09dc • https://ai.googleblog.com/2019/05/efficientnet-improving-accuracy-and.html • https://hackernoon.com/gradient-descent-aynk-7cbe95a778da • http://cdn.aiindex.org/2018/AI%20Index%202018%20Annual%20Report.pdf 81

Recommend


More recommend