Machine Learning: Do legado à vanguarda

A little journey describing what is machine learning, where does it come from, why should we care and how their algorithms look like.

Transcript

2. Olá!
3. Who am I? JUAN RIVILLAS @jprivillaso 1
5. Machine Learning
6. Before that! Let’s talk about some concepts, insights and history 3
7. What is Machine Learning? 4
8. What does learning mean? 5
9. ”Learning is the process of converting experience into expertise/knowledge” 6
10. We thought that only humans and animals were capable of learning 7
12. Neuroscience: Scientific Study of the nervous system 9
13. We looked for a transition From humans to machines 10
14. Machine Learning was introduced by Arthur Samuel in 1959 11
15. One of the pioneers of Artificial Intelligence and Computer Gaming 12
16. ”Give computers the ability to learn without being explicitly programmed” [Samuel A. 1959] 13
17. ”The term machine learning refers to the automated detection of meaningful patterns in data” [ Understanding Machine Learning. From theory to Algorithms, 2014 ] 14
18. We’re get used to 15
19. Let’s change the chip and think slightly different 16
20. Why did it took so long to be hype? 17
21. Storage Speed Performance The keys of evolution 18
22. Artificial Intelligence Machine Learning ? 19
23. Artificial Intelligence Try to build automated Imitation of intelligent behavior [ Understanding Machine Learning. From theory to Algorithms, 2014 ] 20
24. Machine Learning Use the strengths and special abilities of computers to complement human intelligence [ Understanding Machine Learning. From theory to Algorithms, 2014 ] 21
25. Why Machine Learning? 22
26. Most of the problems we face in Machine Learning are related to scenarios that we are familiar with. 23
27. [ Google Trends ] 24
28. Google now is MACHINE LEARNING FIRST 25
31. Be SMARTer 28
32. Understand the behavior of your users Deliver a better service. A customized product. 29
37. Etc.. 34
38. Types of Learning
42. Machine Learning Supervised Semi-supervised Unsupervised Reinforcement 38
43. Machine Learning Supervised Semi-supervised Unsupervised Reinforcement 39
45. Fundamental Question 1 Where does Math appear in Machine Learning? 41
46. Remember Maths Linear Algebra Statistics 42
47. What do I have? Data Data follow a distribution of probability It’s better to work with matrices than pure equations 43
48. Do I have to understand all of these? It’s better, yes! But not strictly necessary 44
49. The most important thing Don’t be afraid of mathematical formalism 45
50. Back to the problem: Collect data Collect Data 46
51. Collect Data Collect data is only the first step 47
52. Collect Data Is the data ready to be used? 48
53. Collect Data - Normalization - Delete useless data 49
54. + Data + Accurracy Collect Data 50
55. What is a model? Fundamental Question 2 51
56. A model can be a function that represent our data = 2 52
57. How to choose a model? Spam Detector Bait Shyness - Rats Learning to Avoid Poisonous Baits 53
58. Spam Detector Spam Not Spam 54
59. Bait Shyness Rats Learning to Avoid Poisonous Baits > Smell > Taste > Consequence after eating? 55
60. It depends! Depends on the problem Conclusion 56
61. Train I already collected data I chose the model I can train my algorithm 57
62. Train What does it mean to train the algorithm? Fundamental Question 2 58
63. Train KNN: K nearest neighbor Lazy Learning Understand the intuition 59
64. Train In this case, the algorithm only stores the data 60
65. Predict Finally, we can Predict =) 61
67. Predict KNN is an algorithm that can be applied to supervised and unsupervised learning 62
68. Predict 63
69. Predict K is my point of reference We will calculate the distance between our point and all the other points 64
70. Predict How to choose K? Validation process! 65
71. Predict Euclidean Distance 66
72. And this is just the beggining of the journey 67
73. Qual linguagem usar? Python R Matlab Java Javascript I would say any language! 68
74. Where to learn? - Coursera: Andrew Ng - Udacity - Udemy 69
77. ML.js www.burakkanber.com/tag/ml-in-js/ github.com/lazyprogrammer References 72
78. Thank you Gracias Obrigado =) @jprivillaso 73
79. Want to work with us? 74 Ask me!