Core ML: A whistlestop tour

Ddd6d3bac7772fa67fc5e312a18bdaec?s=47 sammyd
March 23, 2018

Core ML: A whistlestop tour

Core ML is a framework that makes is super-easy to query machine learning models on Apple devices. But what is machine learning? Let's find out through the art of badly drawn diagrams.

These slides accompany the talk presented at iOSCon 2018 in London.

Ddd6d3bac7772fa67fc5e312a18bdaec?s=128

sammyd

March 23, 2018
Tweet

Transcript

  1. Core ML: a whistlestop tour

  2. Core ML

  3. With Core ML, you can integrate trained machine learning models

    into your app. — developer.apple.com
  4. None
  5. None
  6. classification tell me which of these groups the input falls

    into
  7. regression estimate an output value given this input

  8. clustering how many different classes does the data fall into?

  9. 1 collect data

  10. 2 ground truth

  11. 3 feature vectors

  12. 4 train model

  13. 5 test model

  14. 6 repeat 3, 4 & 5

  15. 7 deploy model

  16. 8 perform prediction

  17. Machine Learning 1. Collect data 2. Ground truth 3. Feature

    vectors 4. Train model 5. Test model 6. Repeat 3, 4 & 5 7. Deploy model 8. Perform prediction
  18. Machine Learning 1. Collect data 2. Ground truth 3. Feature

    vectors 4. Train model 5. Test model 6. Repeat 3, 4 & 5 7. Deploy model 8. Perform prediction
  19. Linear Regression Support Vector Machine Tree Ensembles Bayesian Networks Neural

    Networks
  20. Core ML provides an abstraction across different model types

  21. model interchange format

  22. device optimised implementation

  23. consistent API

  24. that just leaves the hard part

  25. SVM support vector machine

  26. None
  27. None
  28. None
  29. None
  30. None
  31. None
  32. None
  33. None
  34. None
  35. None
  36. None
  37. None
  38. None
  39. example: sentiment analysis

  40. bag of words

  41. demo

  42. convolutional neural networks

  43. None
  44. None
  45. None
  46. None
  47. None
  48. None
  49. None
  50. None
  51. None
  52. None
  53. None
  54. None
  55. None
  56. example: salient object subitizing

  57. demo

  58. conclusion

  59. Core ML is a small part of the story

  60. useful nonetheless

  61. format optimisation API

  62. potentially exciting times

  63. github.com/sammyd /iOSCon2018_WhistlestopCoreML - @iwantmyrealname