Skip to content

cheng-tan/icml2019

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

26 Commits
 
 
 
 

Repository files navigation

Real World Reinforcement Learning Workshop

ICML 2019

ICML session page
Sunday June 9, 2019
2PM - 6:30PM
Room 104

Abstract

Microsoft recently announced the Azure Cognitive Service, Personalizer, aimed at democratizing real world reinforcement learning for context personalization. Its goal is to make reinforcement learning accessible to everyone, not just machine learning experts. Personalizer is the result of a successful partnership between Microsoft Research and Azure Cognitive Services aimed at rapid technology transfer and innovation.

In this workshop you will learn the theory behind contextual bandits and how this applies to content personalization. We will walk you through setting up the service, writing your first application, and optimizing the policy using offline optimization.

Schedule

2PM - 3PM Introduction to reinforcement learning and contextual bandits (Slides)
3PM - 3:30PM Overview of Azure Cognitive Services Personalizer (Slides)
3:30PM - 4PM Break
4PM - 4:30PM Hands on: Setting up SDK and writing first application (Slides)
4:30PM - 5PM Hands on: Counterfactual evaluation and offline policy optimization
5PM - 5:30PM Wrap-up and Q&A

Workshop Instructions

  1. Create free Azure/Microsoft account
  2. Provision free instance of Personalizer
    1. Go to Azure Portal
    2. Search for "Cogitive Services"
    3. On Cognitive Services page click "Add"
    4. Search for "Personalizer (Preview)" and click "Create"
  3. Install Python Client
    1. Download client wheel package
    2. pip install azure_cognitiveservices_personalizer-0.2.0-py2.py3-none-any.whl
  4. Paste your endpoint into line 34 of ./demo/demo.py
  5. Paste your key into line 35 of ./demo/demo.py
  6. run python ./demo/demo.py

Next steps

Presenters

  • John Langford
  • Rodrigo Kumpera
  • Alexey Taymanov
  • Cheng Tan

Releases

No releases published

Packages

No packages published