1. m Machine Learning F# and Accord.net

2. Alena Dzenisenka Software architect at Luxoft Poland Member of F# Software Foundation Board of Trustees Researcher in the field of mathematical theoretical possible in modern programming concepts Speaker and active software engineering community member @ lenadroid

3. Machine Learning

4. Why machine learning? What is the data? How? Questions

5. Data Questions.

6. Data reality :\

7. Path to grasping machine learning and data science…

8. Contents Multiple Linear  Regression Logistic Regression  Classification Neural Networks  Classification K Means  Clustering What’s next?

10. F# for machine learning and data science!

11. Linear Regression

12. MATH

13. How to predict? 1.Make a guess. 2.Measure how wrong the guess is. 3.Fix the error.

14. Make a guess!

15. Make a guess? What does it mean?... Hypothesis /guess : weights

16. Find out our mistake… Cost function/ Mistake function: … and minimize it:

17. Mistake function looks like… Global minimums

18. How to reduce the mistake? Update each slope parameter until Mistake Function minimum is reached: Simultaneously Alpha Learning rate Derivative Direction of moving

19. Fix the error

20. Multiple Linear Regression X [ ] – Predictors: Statistical data about bike rentals for previous years or months. Y – Output: Amount of bike rentals we should expect today or some other day in the future. * Y is not nominal, here it’s numerical continuous range.

21. Make a guess!

22. Fix the error

23. Multiple linear regression: Bike rentals demand “Talk is cheap. Show me the code.”

24. What to remember? 1.Simplest regression algorithm 2.Very fast, runs in constant time 3.Good at numerical data with lots of features 4. Output from numerical continuous range 5. Linear hypothesis 6. Uses gradient descent Linear Regression

25. Logistic Regression

26. Hypothesis function Estimated probability that Y = 1 on input X

27. Mistake function Mistake function is the cost for a single training data example h(x)

28. Full mistake function 1. Uses the principle of maximum likelihood estimation. 2. We minimize it same way as with Linear Regression

29. “Talk is cheap. Show me the code.” Logistic Regression Classification Example

30. What to remember? Classification algorithm Classification algorithm Output is the binary value, either 1 or 0 Output is the binary value, either 1 or 0 Relatively small number of predictors Relatively small number of predictors Uses logistics function for hypothesis Uses logistics function for hypothesis Has the cost function that is convex Has the cost function that is convex Uses gradient descent for correcting the mistake Uses gradient descent for correcting the mistake Logistic Regression

31. At this point…

32. Machine Learning What society thinks I do…What other programmers think I do…

33. What I really do is…

34. Neural Networks

35. Neuron

36. Neural Networks

37. Layer transitions! InputWeightsActivation function Activation unit (value which is computed and outputted by the node)

38. Activation function? Step activation function Used in binary classification Sigmoid activation function Uses logistic function to perform the activation

39. Forward propagation x x x x a a a a a a a X = X X = a 1 X = a 2 Hypothetic Outputs! Activation Computational Nodes! Predictors

40. Multi class? OR x x x a a a a a x x x a a a a a

41. Cost function? x x x Looks difficult? No, it’s really not so complicated… 1.Outputs the vector with size == number of output classes 2.Similar to Logistic Regression cost function 3.Finds the mistake of our hypothesis and real outputs

42. Minimize the error! x x x x Error value And then… recalculate weights!

43. “Talk is cheap. Show me the code.” Neural Networks Classification Example.

44. What to remember? Extremely powerful Extremely powerful Slow learning / Fast prediction Slow learning / Fast prediction Big number of predictors Big number of predictors Complex hierarchical relationships Complex hierarchical relationships No need to understand the data No need to understand the data Wide applications Wide applications Neural Networks

45. K-Means

46. Clustering

47. Birth-death percentage rates

48. … with clustering K = 2

49. … with clustering K = 3

50. … with clustering K = 4

51. What’s next?

53. I’m Lena @lenadroid

54. Thank you!