Machine Learning for Developers

Devoxx Belgium, Antwerp, November 10th, 2016

Have you always wanted to add predictive capabilities or voice recognition to your application, but haven’t been able to find the time or the right technology to get started? Everybody wants to build smart apps, but only a few are Data Scientists. We had the same issue inside Amazon, so we created a Machine Learning engine that Developers can easily use. The same approach is now available in the AWS cloud. And we introduced Amazon Alexa to build engaging voice experiences for your services and devices: if you are a device maker, and your connected product has a microphone and a speaker, the Alexa Voice Service (AVS) enables you to add voice-powered experiences to your connected devices. And you can also use the Alexa Skills Kit (ASK) to teach new skills to Alexa!

Transcript

1. Machine Learning
   for Developers
   Danilo Poccia
   Technical Evangelist, AWS
   @danilop
   Sébastien Stormacq
   Solution Architect, Alexa
   @sebsto
   

   View Slide

2. What to Expect from this Session
   Understand Machine Learning
   Terminology & Challenges
   Implement a Machine Learning Model
   Add Predictive Capabilities to your App
   Provide Your Customer with Voice UX
   

   View Slide

3. Credit: Gerry Cranham/Fox Photos/Getty Images
   http://www.telegraph.co.uk/travel/destinations/europe/united-kingdom/england/london/galleries/The-history-of-the-Tube-in-pictures-150-years-of-London-Underground/1939-ticket-examin/
   

   View Slide

4. Credit: Gerry Cranham/Fox Photos/Getty Images
   http://www.telegraph.co.uk/travel/destinations/europe/united-kingdom/england/london/galleries/The-history-of-the-Tube-in-pictures-150-years-of-London-Underground/1939-ticket-examin/
   1939 London Underground
   

   View Slide

5. Batch
   Report
   

   View Slide

6. Batch
   Report
   Real-time
   Alerts
   

   View Slide

7. Batch
   Report
   Real-time
   Alerts
   Prediction
   Forecast
   

   View Slide

8. Predictions
   

   View Slide

9. Data Predictions
   

   View Slide

10. Model
    Data Predictions
    

    View Slide

11. Model
    

    View Slide

12. Machine Learning
    

    View Slide

13. Supervised
    Learning
    Machine Learning
    Unsupervised
    Learning
    The task of inferring
    a model
    from labeled
    training data
    The task of inferring
    a model
    to describe
    hidden structure
    from unlabeled data
    

    View Slide

14. Reinforcement
    Learning
    Perform
    a certain goal in a
    dynamic environment,
    without an explicit
    “teacher”
    

    View Slide

15. Driving a vehicle
    Playing a game

    against an opponent
    R
    einforcem
    ent
    Learning
    

    View Slide

16. Clustering
    U
    nsupervised

    Learning
    

    View Slide

17. Clustering
    U
    nsupervised

    Learning
    

    View Slide

18. Clustering
    U
    nsupervised

    Learning
    

    View Slide

19. Regression
    Binary Classification
    Multi-class Classification
    Supervised

    Learning
    

    View Slide

20. Validation
    Supervised

    Learning
    

    View Slide

21. Training from Labeled Data
    Supervised

    Learning
    Training
    Validation
    70%
    30%
    

    View Slide

22. Training the Model
    Minimizing the Error Function
    Supervised

    Learning
    

    View Slide

23. Be Careful of Overfitting
    Supervised

    Learning
    

    View Slide

24. Be Careful of Overfitting
    Supervised

    Learning
    

    View Slide

25. Be Careful of Overfitting
    Supervised

    Learning
    

    View Slide

26. Better Model,
    Different Predictions
    Supervised

    Learning
    

    View Slide

27. Supervised

    Learning
    Better Model,
    Different Predictions
    

    View Slide

28. Regularization
    Adding a “cost” for using

    large parameters in the model
    L1, L2
    Supervised

    Learning
    

    View Slide

29. ?
    Data Model
    

    View Slide

30. Amazon EMR
    with Spark (MLib)
    Data Model
    

    View Slide

31. Data Scientists
    “Scalability”
    

    View Slide

32. Amazon

    Machine Learning

    (Amazon ML)
    Data Model
    

    View Slide

33. Amazon

    Machine Learning

    (Amazon ML)
    Data Model
    Batch
    Predictions
    

    View Slide

34. Amazon

    Machine Learning

    (Amazon ML)
    Data Model
    Batch
    Predictions
    Real-time
    Predictions
    

    View Slide

35. Binary Classification
    Multiclass
    Classification
    Regression
    Logistic Regression

    (Logistic Loss
    Function + SGD)
    Multinomial Logistic
    Regression

    (Multinomial Logistic
    Loss + SGD)
    Linear Regression

    (Squared Loss
    Function + SGD)
    The optimization technique used in Amazon ML is

    online Stochastic Gradient Descent (SGD)
    

    View Slide

36. 
    ...
    
    

    View Slide

37. 
    Bike Sharing
    
    

    View Slide

38. View Slide

39. View Slide

40. All Users
    

    View Slide

41. All Users
    Casual Users
    Registered Users
    

    View Slide

42. 
    ...
    
    

    View Slide

43. What about Deep Learning?
    

    View Slide

44. Neural Networks
    Perceptron
    Layers
    

    View Slide

45. Perceptron
    https://upload.wikimedia.org/wikipedia/commons/8/8c/Perceptron_moj.png
    https://upload.wikimedia.org/wikipedia/commons/thumb/f/f1/Logistic-sigmoid-vs-scaled-probit.svg/240px-Logistic-sigmoid-vs-scaled-probit.svg.png
    

    View Slide

46. Neural
    Network
    Architectures
    http://www.asimovinstitute.org/neural-network-zoo/
    

    View Slide

47. Deep Learning AMI
    5 Deep Learning Frameworks
    MXNet, Caffe, Tensorflow, Theano, and Torch
    Pre-installed components to speed productivity, such
    as Nvidia drivers, cuDNN, Anaconda, Python 2 & 3
    AWS Integration
    R
    eady
    to
    use
    on
    A
    m
    azon
    EC
    2
    

    View Slide

48. Amazon EC2 P2 Instances
    Up to:
    • 16 NVIDIA K80 GPUs
    • 64 vCPUs 732 GiB of host memory
    • combined 192 GB of GPU memory
    • 40 thousand parallel processing cores
    • 70 teraflops (single precision)
    • over 23 teraflops (double precision).
    • GPUDirect™ for up to 16 GPUs
    G
    PU
    Instances
    

    View Slide

49. What about interacting with devices
    in a more intuitive way, using voice?
    

    View Slide

50. View Slide

51. 
    ...
    
    

    View Slide

52. Create Great Content:
    ASK is how you connect

    to your consumer
    THE ALEXA ECOSYSTEM
    Supported by two powerful frameworks
    A L E X A 

    V O I C E 

    S E R V I C E
    Unparalleled Distribution:
    AVS allows your content

    to be everywhere
    Lives In The Cloud
    Automated Speech Recognition
    (ASR)
    Natural Language Understanding
    (NLU)
    Always Learning
    A L E X A 

    S K I L L S 

    K I T
    

    View Slide

53. UNDER THE HOOD OF ASK
    A closer look at how the Alexa Skills Kit process a request and returns an appropriate
    response
    You Pass Back a Textual
    or Audio Response
    You Pass Back a Graphical Response
    Alexa Converts Text-to-Speech (TTS) &
    Renders Graphical Component
    Respond to Intent
    through Text & Visual
    Alexa sends Customer
    Intent to Your Service
    User Makes
    a Request Alexa Identifies Skill & Recognizes
    Intent Through ASR & NLU
    Your Service
    processes
    Request
    Audio Stream is
    sent up to Alexa
    

    View Slide

54. ALEXA SKILL KIT
    High Level Overview
    Your Code
    

    View Slide

55. ALEXA SKILL KIT
    High Level Overview
    Amazon EC2
    

    View Slide

56. ALEXA SKILL KIT
    High Level Overview
    Availability Zone 1
    Web tier
    App tier
    RDS
    (Master)
    Availability Zone 2 RDS
    (Standby)
    

    View Slide

57. ALEXA SKILL KIT
    High Level Overview
    Elastic Beanstalk environment
    Auto Scaling group
    Elastic Beanstalk container
    Auto Scaling group
    Elastic Beanstalk container
    Prod 1
    Prod 2
    Route 53
    

    View Slide

58. ALEXA SKILL KIT
    High Level Overview
    AWS Lambda
    

    View Slide

59. ALEXA SKILL KIT
    Voice Model
    Intents
    Utterances
    Slots
    

    View Slide

60. ALEXA SKILL KIT
    Voice Model
    Intents
    Utterances
    Slots
    

    View Slide

61. ALEXA SKILL KIT
    Voice Model
    Intents
    Utterances
    Slots
    

    View Slide

62. ALEXA SKILL KIT
    Voice Model
    Intents
    Utterances
    Slots
    

    View Slide

63. ALEXA SKILL KIT
    Typescript Code
    

    View Slide

64. ALEXA SKILL KIT
    Alexa App
    

    View Slide

65. 
    ...
    
    

    View Slide

66. Your Skill
    (Lambda function)
    Amazon
    Machine Learning
    get real-time predictions
    invoke
    Weather
    Forecast
    Historical Data
    get forecast
    build & train model
    B
    ike
    Sharing
    D
    em
    o
    A
    rchitecture
    

    View Slide

67. Build Apps With Services,
    Not Servers
    

    View Slide

68. aws.amazon.com/free
    developers.amazon.com
    

    View Slide

69. Thank you!
    @danilop @sebsto
    

    View Slide