Scalable and Robust Representation Learning in Large output Spaces

  • Babbar, Rohit (Principal investigator)
  • Ullah, Nasib (Project Member)

Project Details


Artificial Intelligence deals with automatically learning underlying pattern from observational data. The task of a spam detector for emails is to make a choice between two possibilities - spam vs non-spam. Imagine tagging a Wikipedia page, which has more than a million categories to choose from. Would it not be nice to able to assign relevant categories to Wikipedia pages automatically, or predicting top relevant items when shopping on Amazon or show you the most relevant tweets or Facebook posts on top, out of thousands of items/tweets/posts. This research project attempts to overcome the following challenges when dealing with such scenarios consisting of - (i) very large number of choices, (ii) incomplete observations of choices and (iii) scarce observational data per choice. The goal of this project is to build computer programs which help in personalised, accurate and diverse recommendation of relevant items/choices to users in real-time for a richer experience.
Effective start/end date01/09/202231/08/2026

Collaborative partners


Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.