sagemaker-rec-demo

Building a Recommender System with Amazon SageMaker Factorization Machines and BlazingText

Contents

Notebook: fm_amazon_recommender.ipynb

Main Workshop:

• Background (Factorization Machines)
• Setup
  • Spin up SageMaker hosted notebook instance in console
  • Add SageMaker IAM policy to this SageMaker notebook to allow S3 read/write access
  • Create new S3 bucket (first cell)
  • Import necessary libraries (second cell)
• Dataset
  • Overview
  • Source: Amazon Reviews Public Dataset
  • Features
• Data preprocessing
• Training
  • Create SageMaker estimator
  • Launch training job
• Host
  • Deploy endpoint to perform inference

Extra Credit:

• Background (New challenges, word2vec, BlazingText)
• Data Augmentation
• Dimensionality reduction with t-SNE
• Train
• Host