/AIHackKrakow

Primary LanguageRoffMIT LicenseMIT

Introduction

In this course, you learn how to create an end-to-end data science solution, applying advanced machine learning (ML) approaches to a real-world scenario, variants of which can be found across industry verticals. This hands-on training covers various important machine learning algorithms. Use notebooks to understand the math behind data science and learn best practices for cleansing and manipulating your data to gather insights from it. Then, build an ML model on this data using techniques that will give you a foray into a data scientist’s work. Further, see MLOps (DevOps for ML) in action while learning how to productionize your models. Learn how to do the following: (1) perform data preparation and feature engineering with Pandas dataframes; (2) conduct model development with the Scikit-Learn ML library; (3) learn essentials of machine learning experimentation (model management and evaluation) with AML service; (4) perform tuning of hyperparameters with HyperDrive on AML Compute; (5) quickly find the best combination of ML algorithm and feature selection with automated machine learning; and (6) set up real-time scoring with Azure Kubernetes Services (AKS).

Agenda

time length format presenter topic
09:00 - 09:15 15 min OPEN both Introducing instructors and going over course objectives
09:15 - 09:30 15 min SLIDES Alexandre Vilcek Rquirements for an advanced ML platform
09:30 - 10:00 30 min LAB Preparing lab environment and intro to Jupyter
10:00 - 10:15 15 min Break
10:15 - 10:30 15 min OPEN Alexandre Vilcek What we learn in this notebook
10:30 - 11:00 30 min LAB Training an ML model using sklearn
11:00 - 11:15 15 min PORTAL Alexandre Vilcek Wrap up and go over slides
11:15 - 11:45 15 min SLIDES Seth Mottaghinejad How Azure offers a comprehensive AI/ML platform
11:45 - 12:00 15 min Q&A
12:00 - 13:00 60 min Lunch
13:00 - 13:15 15 min SLIDES Seth Mottaghinejad Introduce hyperdrive concepts
13:15 - 13:45 30 min LAB Run through hyperdrive notebook
13:45 - 14:00 15 min PORTAL Seth Mottaghinejad Wrap up and go over slides
14:00 - 14:15 15 min SLIDES Alexandre Vilcek Introduce AutoML and explainability concepts
14:15 - 14:45 30 min LAB Run through AutoML and explainability notebook
14:45 - 15:00 15 min PORTAL Alexandre Vilcek Wrap up and go over slides
15:00 - 15:15 15 min Break
15:15 - 15:30 15 min SLIDES Seth Mottaghinejad Introduce deployment concepts
15:30 - 16:15 45 min LAB Run through deployment notebook
16:15 - 16:30 15 min PORTAL Seth Mottaghinejad Wrap up deployment
16:30 - 17:00 30 min Wrap up and Q&A

Lab setup

  1. Log into https://aka.ms/ignite-PRE12 using activation code IGNITE5301 and sign up to obtain your lab credentials.
  2. Log into jupyter lab and follow instructor's quick introduction of Jupyter lab.
  • types of notebook cells
  • keyboard shortcuts
  • kernels and conda
  1. Open a terminal window from Juyter by clicking on New > Terminal and type the following command to clone the course repository: 
    cd /data/home/labuser/notebooks
git clone https://github.com/Azure/Ignite2019-pre-day.git # clone the course repo
  2. If you return to the Juypter file browser, you should now see a folder with the course content (if not refresh the page). Course folder
  3. Navigate to the Ignite2019-pre-day folder and open the notebook 00_Lab_setup.ipynb and run through the steps in the notebook. Open notebook
  4. Restart the data science virtual machine from the Azure portal by
    • log into the Azure portal using the credentials in the lab credentials page
    • click on Resource groups and click on your resource group name (should be something like ODL-machine-learning-######)
    • click on the virtual machine resource (should be named something like dsvm-#####), and Resource group
    • click on Restart Restart DSVM