dsi-values

An online movie recommendation site for Charlottesville Virginia using Contextual Bandit (CB) algorithms/models.

Algorithms

• Linear CB modification of Abbeel and Ngs "Apprenticeship Learning via Inverse Reinforcement Learning".
• Kernel CB modification of Abbeel and Ngs "Apprenticeship Learning via Inverse Reinforcement Learning".

Algorithm Implmenetation

• Algorithms are implemented in JS and run in the client's web-browser.
• All vector and matrix calculations are done utilizing TensorFlow.js.

Hosting

• The HTML and JS files for the site are stored on S3 and use AWS's CDN, DNS and Certificate Manager for requests.
• The live version of this project is deployed at https://dsi.markrucker.net

Data

• The backend of this site runs a nightly job to query the next week's movies from:
  • http://developer.tmsapi.com/
  • http://www.omdbapi.com/
• The nightly backend job stores the movie data in AWS's No-SQL DynamoDB
• The frontend of this site uses DynamoDB's JS API to directly query the movie times and theaters