Collaborative filtering with maximum entropy

Dmitry Pavlov, Eren Manavoglu, C. Lee Giles, David M. Pennock

Research output: Contribution to journalReview articlepeer-review

18 Scopus citations

Abstract

The collaborative filtering technique with maximum entropy, or maxnet, approach for generating online recommendations is discussed. Maxnet enables fast model querying and performs better than its competitors in terms of accuracy. Collaborative filtering methods work by assessing the similarities among users on the basis of their overlap in document requests, then recommending to a given user documents that like-minded users access previously. The model-based approach provides an attractive alternative to current recommenders in terms of the quality of its predictions.

Original languageEnglish (US)
Pages (from-to)40-48
Number of pages9
JournalIEEE Intelligent Systems
Volume19
Issue number6
DOIs
StatePublished - Nov 2004
Externally publishedYes

All Science Journal Classification (ASJC) codes

  • Computer Networks and Communications
  • Artificial Intelligence

Fingerprint Dive into the research topics of 'Collaborative filtering with maximum entropy'. Together they form a unique fingerprint.

Cite this