How can Azure superpower your data science journey?

How can Azure superpower your data science journey?

This deck serves as an introduction to Azure Machine learning services.
it walks you through a brief intro of ML, Azure Machine learning services, Distributed hyperparameter tuning and Auto ML

Ecdea9b9714877b86cee08458f085481?s=128

Tania Allard

November 30, 2019
Tweet

Transcript

  1. How can Azure superpower your Data Science journey? UCL data

    science hackathon 2019-11-30 Tania Allard, PhD ixek
  2. About me Developer advocate @ Microsoft – scientific computing and

    machine learning Champion for open source in research and education Champion for diversity, inclusion and accesible tech PSF fellow GDE for Tensorflow
  3. Outline 3 Azure Machine learning A 101 on using machine

    learning to solve problems Super quick intro to ML Getting started with AML Advanced features for ML Speeding up 1 2 3
  4. What is machine learning? Quick 101

  5. What is AI, ML, and DL? Artificial intelligence (AI) is

    a technique that enables computers to mimic human intelligence. It includes machine learning. Machine learning (ML) is a subset of artificial intelligence that includes techniques (such as deep learning) that enable machines to improve at tasks with experience. Deep learning (DL) is a subset of machine learning based on artificial neural networks that permit a machine to train itself.
  6. Traditional programming Machine learning

  7. Machine learning

  8. Machine learning FINDING (AND EXPLOITING) PATTERNS IN DATA REPLACING “HUMAN

    WRITING CODE” WITH “HUMAN SUPPLYING DATA” STARTS WITH A SHARP QUESTION
  9. When is ML the right tool? Start by asking a

    question: What are the forecasted sales quantities per item in the next 4 weeks?
  10. When should you use machine learning? • Regression: how much

    / how many • Classification: which class does it belong to? • Clustering: are there different groups? Which does it belong to? • Anomaly Detection: is this weird? • Recommendation: which option should I choose? supervised learning unsupervised learning
  11. https://cda.ms/17L

  12. How can Azure help you to develop ML solutions?

  13. Azure Bot Service Azure Cognitive Services Azure Cognitive Search Azure

    Machine Learning Knowledge mining AI apps & agents Machine learning Azure AI
  14. Machine Learning on Azure Azure Machine Learning Domain Specific Pretrained

    Models To reduce time to market Azure Databricks Machine Learning VMs Popular Frameworks To build machine learning and deep learning solutions TensorFlow PyTorch ONNX Azure Machine Learning Language Speech … Search Vision Productive Services To empower data science and development teams Powerful Hardware To accelerate deep learning Scikit-Learn PyCharm Jupyter Familiar Data Science Tools To simplify model development Visual Studio Code Command line CPU GPU FPGA
  15. Azure Machine Learning Azure Cloud Services Python SDK ü Prepare

    Data ü Build Models ü Train Models ü Manage Models ü Track Experiments ü Deploy Models That enables you to: Cross-Platform CLI
  16. Machine Learning Typical E2E Process Prepare Data Register and Manage

    Model Train & Test Model Build Image Build model (your favorite IDE) Deploy Service Monitor Model Prepare Experiment Deploy Orchestrate …
  17. Azure Machine Learning Datasets – registered, known data sets Experiments

    – Training runs Pipelines – Training workflows Models – Registered, versioned models Endpoints: Real-time Endpoints – Deployed model endpoints Pipeline Endpoints – Training workflow endpoints Compute – Managed compute Environments – defined training and inference environments Datastores – Connections to data
  18. 19 Familiar example: digits recognition

  19. 20 Familiar example: digits recognition

  20. Workspace and compute

  21. Azure Machine Learning Datasets – registered, known data sets Experiments

    – Training runs Pipelines – Training workflows Models – Registered, versioned models Endpoints: Real-time Endpoints – Deployed model endpoints Pipeline Endpoints – Training workflow endpoints Compute – Managed compute Environments – defined training and inference environments Datastores – Connections to data
  22. Your favourite platforms 23

  23. Running experiments

  24. Azure Machine Learning Datasets – registered, known data sets Experiments

    – Training runs Pipelines – Training workflows Models – Registered, versioned models Endpoints: Real-time Endpoints – Deployed model endpoints Pipeline Endpoints – Training workflow endpoints Compute – Managed compute Environments – defined training and inference environments Datastores – Connections to data
  25. None
  26. Deploying your models

  27. Azure Machine Learning Datasets – registered, known data sets Experiments

    – Training runs Pipelines – Training workflows Models – Registered, versioned models Endpoints: Real-time Endpoints – Deployed model endpoints Pipeline Endpoints – Training workflow endpoints Compute – Managed compute Environments – defined training and inference environments Datastores – Connections to data
  28. My Computer Data Store Azure ML Workspace Compute Target

  29. The 8 Azure ML Train Steps

  30. Deploy image Azure Kubernetes Service (AKS) Azure Container Instance

  31. Some advanced AML features

  32. Distributed Hyperparameter Tuning

  33. What are Hyperparameters? { “learning_rate”: uniform(0, 1), “num_layers”: choice(2, 4,

    8) … } Config1= {“learning_rate”: 0.2, “num_layers”: 2, …} Config2= {“learning_rate”: 0.5, “num_layers”: 4, …} Config3= {“learning_rate”: 0.9, “num_layers”: 8, …} …
  34. Typical ‘manual’ approach to hyperparameter tuning Dataset Training Algorithm 1

    Hyperparameter Values – config 1 Model 1 Hyperparameter Values – config 2 Model 2 Hyperparameter Values – config 3 Model 3 Model Training Infrastructure Training Algorithm 2 Hyperparameter Values – config 4 Model 4 Complex Tedious Repetitive Time consuming Expensive
  35. None
  36. None
  37. None
  38. None
  39. Automated ML

  40. Mileage Condition Car brand Year of make Regulations … Parameter

    1 Parameter 2 Parameter 3 Parameter 4 … Gradient Boosted Nearest Neighbors SVM Bayesian Regression LGBM … Mileage Gradient Boosted Criterion Loss Min Samples Split Min Samples Leaf Others Model Which algorithm? Which parameters? Which features? Car brand Year of make Model creation is typically a time-consuming process
  41. Which algorithm? Which parameters? Which features? Mileage Condition Car brand

    Year of make Regulations … Gradient Boosted Nearest Neighbors SGD Bayesian Regression LGBM … Nearest Neighbors Criterion Loss Min Samples Split Min Samples Leaf XYZ Model Iterate Gradient Boosted N Neighbors Weights Metric P ZYX Mileage Car brand Year of make Car brand Year of make Condition Track Model creation is typically a time-consuming process
  42. Track Which algorithm? Which parameters? Which features? Iterate Model creation

    is typically a time-consuming process
  43. Enter data Define goals Apply constraints Output Automated Machine Learning

    accelerates model development Input Intelligently test multiple models in parallel Optimized model
  44. Caution

  45. Auto ML should not be an excuse for black-box ML

  46. None
  47. None
  48. None
  49. Experiment! This is only a taster of what you can

    do. Experiment with multiple scenarios, techniques and products Get your Azure subscription and claim your $100
  50. Useful resources • GitHub Repository : https://github.com/trallard/ML-in-AML • Azure Machine

    learning • Create development environment for Machine learning • Hyperparameter tuning in AML • AML Python SDK • AML Pipelines
  51. Useful resources • Getting started with Auto ML • Intro

    to AML – MS Learn • Automate model select with AML - MS Learn • Train local model with AML - MS Learn
  52. Tania Allard, PhD @ixek 56 Thanks!