Machine Learning on Mobile—a primer

Ddd6d3bac7772fa67fc5e312a18bdaec?s=47 sammyd
October 07, 2018

Machine Learning on Mobile—a primer

Machine learning is incredibly powerful, and with the introduction of Core ML on iOS and ML Kit on Android, finally becoming usable on mobile devices. But what exactly is 'machine learning'.
In this talk you'll hear about the basics of machine learning—the fact that at its simplest level it's just trivial mathematical functions. We'll build up our own models in the talk, using principles from high school mathematics, demonstrating that machine learning isn't as complex as it may seem.
We'll then look at how to work with more advanced models, including convolutional neural networks (CNNs) and how they work in iOS with Core ML.
At the end of the talk you will have a better understanding of exactly what 'machine learning' is, along with some practical knowledge of how to use machine learning in their own apps.

Ddd6d3bac7772fa67fc5e312a18bdaec?s=128

sammyd

October 07, 2018
Tweet

Transcript

  1. MACHINE LEARNING ON MOBILE: A primer @iwantmyrealname

  2. CORE ML @iwantmyrealname

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

    into your app. — developer.apple.com @iwantmyrealname
  4. @iwantmyrealname

  5. @iwantmyrealname

  6. CLASSIFICATION tell me which of these groups the input falls

    into @iwantmyrealname
  7. REGRESSION estimate an output value given this input @iwantmyrealname

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

    @iwantmyrealname
  9. 1 COLLECT DATA @iwantmyrealname

  10. 2 GROUND TRUTH @iwantmyrealname

  11. 3 FEATURE VECTORS @iwantmyrealname

  12. 4 TRAIN MODEL @iwantmyrealname

  13. 5 TEST MODEL @iwantmyrealname

  14. 6 REPEAT 3, 4 & 5 @iwantmyrealname

  15. 7 DEPLOY MODEL @iwantmyrealname

  16. 8 PERFORM PREDICTION @iwantmyrealname

  17. 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 @iwantmyrealname
  18. 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 @iwantmyrealname
  19. 1. Collect data 2. Ground truth 3. Feature vectors 4.

    Train model 5. Test model 6. Repeat 3, 4 & 5 7. Deploy model @iwantmyrealname
  20. LINEAR REGRESSION SUPPORT VECTOR MACHINE Tree Ensembles BAYESIAN NETWORKS Neural

    Networks @iwantmyrealname
  21. CORE ML PROVIDES AN abstraction ACROSS different MODEL TYPES @iwantmyrealname

  22. MODEL INTERCHANGE format @iwantmyrealname

  23. DEVICE OPTIMISED implementation @iwantmyrealname

  24. CONSISTENT API @iwantmyrealname

  25. THAT JUST LEAVES THE hard PART @iwantmyrealname

  26. SVM SUPPORT VECTOR MACHINE @iwantmyrealname

  27. @iwantmyrealname

  28. @iwantmyrealname

  29. @iwantmyrealname

  30. @iwantmyrealname

  31. @iwantmyrealname

  32. @iwantmyrealname

  33. @iwantmyrealname

  34. @iwantmyrealname

  35. @iwantmyrealname

  36. @iwantmyrealname

  37. @iwantmyrealname

  38. @iwantmyrealname

  39. @iwantmyrealname

  40. example: SENTIMENT ANALYSIS @iwantmyrealname

  41. bag OF words @iwantmyrealname

  42. DEMO @iwantmyrealname

  43. CONVOLUTIONAL NEURAL NETWORKS @iwantmyrealname

  44. @iwantmyrealname

  45. @iwantmyrealname

  46. @iwantmyrealname

  47. @iwantmyrealname

  48. @iwantmyrealname

  49. @iwantmyrealname

  50. @iwantmyrealname

  51. @iwantmyrealname

  52. @iwantmyrealname

  53. @iwantmyrealname

  54. @iwantmyrealname

  55. @iwantmyrealname

  56. @iwantmyrealname

  57. example: SALIENT OBJECT SUBITIZING @iwantmyrealname

  58. DEMO @iwantmyrealname

  59. Create ML @iwantmyrealname

  60. macOS FRAMEWORK @iwantmyrealname

  61. LIMITED but POWERFUL @iwantmyrealname

  62. DEMO @iwantmyrealname

  63. CONCLUSION @iwantmyrealname

  64. Core ML IS A small PART OF THE STORY @iwantmyrealname

  65. useful NONETHELESS @iwantmyrealname

  66. FORMAT optimisation API @iwantmyrealname

  67. POTENTIALLY exciting TIMES @iwantmyrealname

  68. @iwantmyrealname

  69. github.com/sammyd /MOBILEOPTIMISED2018_ML SAM@RAZEWARE.COM @IWANTMYREALNAME @iwantmyrealname