Fast Algorithms for Recommender Systems

Date
2017-04-11
Authors
Agnihotri, Souparni
Major Professor
Advisor
Committee Member
Journal Title
Journal ISSN
Volume Title
Publisher
Altmetrics
Authors
Research Projects
Organizational Units
Journal Issue
Department
Electrical and Computer Engineering
Abstract

Recommender systems are information filtering systems that deal with the problem of information overload by filtering vital information from a large amount of data retrieved based on user preferences, interests or observed behavior about an item. The challenge is to come up with a fast and efficient algorithm that predicts user preferences on unseen items using a combination of content-based filtering (leveraging metadata about the users/items) and collaborative filtering (leveraging other user ratings). The algorithm uses a parallel computing framework called GPUFish to solve very large scale matrix completion problems. It addresses the main issue of predicting future ratings on a certain product, given a subset of user ratings. Such an algorithm is beneficial for large-scale content providers like Netflix and Amazon to filter their data in a fast and efficient manner and get accurate results. The testing of this algorithm is done through a food application which is implemented using JavaScript and a MongoDB database. The application prompts users to rate different food recipes and this information is used for predicting user ratings in the future. I am involved in developing the android application that uses the Volley networking library to interact with the database and have designed multiple screens to ease the process of user information retrieval.

Comments
Description
Keywords
Citation
DOI
Source