Getting Started with Machine Learning

Transcript

1. Getting Started with Machine Learning Mike Fowler - Senior Site

   Reliability Engineer - Public Cloud Practice PLACE CUSTOMER LOGO HERE

2. • What is Machine Learning? • The AWS Machine Learning

   Stack • ML Use Cases • Machine Learning: The Forgotten Service • SageMaker Agenda

3. London PostgreSQL Meetup January 24th 2019 About Me

4. None

5. Ethics Source: https://peakcare.wordpress.com/2011/10/05/heads-in-the-sand/

6. What is … Machine Learning?

7. How do Machines Learn? Source: https://towardsdatascience.com/machine-learning-types-2-c1291d4f04b1

8. Machine Learning Concepts • Models - Mathematical equation with a

   solution space approximating the outputs for the given inputs • Feature Engineering - Process of identifying & creating features from the data that will influence/assist the model • Training - Repeated process attempting to find the model that is “just right” such that it does not overfit or underfit the training data

9. Dang it Jim, I’m an Engineer not a Mathematician!

10. The AWS Machine Learning Stack

11. Use Case: Audio Description for Images

12. Use Case: Audio Description for Images Lambda

13. Use Case: Audio Description for Images Rekognition Image Lambda

14. Use Case: Audio Description for Images Polly Rekognition Image Lambda

15. Use Case: Corporate Updates For All

16. Use Case: Corporate Updates For All Transcribe

17. Use Case: Corporate Updates For All Transcribe Translate

18. Use Case: Corporate Updates For All Transcribe Translate Polly

19. Use Case: Corporate Updates For All Transcribe Translate Polly S3

20. Use Case: Corporate Updates For All Transcribe Translate Polly S3

    CloudFront

21. The Forgotten Service

22. Identify a Problem to Solve Many PagerDuty incidents resolve before

    I respond disrupting my sleep needlessly

23. Identify a Problem to Solve Many PagerDuty incidents resolve before

    I respond disrupting my sleep needlessly

24. Source Relevant Data

25. Input Data

26. Input Data

27. Target

28. Target

29. Target

30. Train the Model

31. The Lambda Architecture Master Data Serving Layer Batch Layer Speed

    Layer S3 EMR Kinesis Streams Glue Redshift (Batched Views) DynamoDB (Real-Time Views)

32. ML Model The Lambda Architecture + ML Master Data Serving

    Layer Batch Layer Speed Layer S3 EMR Kinesis Streams Glue Redshift (Batched Views) DynamoDB (Real-Time Views) Amazon Machine Learning

33. SageMaker

34. Feature Engineering • Most models only take numeric input •

    Values often need to be constrained - Scale Min/Max - Logarithm • Some values can’t be used - Identifiers - Attributes that wouldn’t be known when making a prediction

35. SageMaker

36. SageMaker

37. Make Predictions

38. Mike Fowler mlfowler Questions ? gh-mlfowler mlfowler_ www.mlfowler.com mike.fowler@claranet.uk

39. None