Contextual Bandits with Amazon SageMaker RL

This Notebook demonstrates how you can manage your own contextual multi-armed bandit workflow on SageMaker using the built-in Vowpal Wabbit (VW) container to train and deploy contextual bandit models. We show how to train these models that interact with a live environment (using a simulated client application) and continuously update the model with efficient exploration.

Contents

• bandits_statlog_vw_custom.ipynb: Notebook used for running the contextual bandit notebook.
  
• config.yaml: The configuration parameters used for the bandit example.
  
• sim_app: Simulated client application that pings SageMaker for recommended action given a state. Also computes the rewards for each interaction.
  
• common: Code that manages the different AWS components required for training workflow.
  
• src:
  • train-vw.py: Script for training with Vowpal Wabbit library.
  • eval-cfa-vw.py: Script for evaluation with Vowpal Wabbit library.