Eigentaste: An Efficient Algorithm For Recommender Systems
- Technology Benefits
- Efficient and effective personalized information retrieval. Multiple types of products, services, or data without customization for each type.Avoids semantic categories byrelying solely upon numerical ratings.Each content block and each user is treated as a"black box" to which statistical pattern recognition techniques are applied.
- Technology Application
- Recommendation of diverse content, such as books, movies, toys, stocks, and music.
- Detailed Technology Description
- None
- Supplementary Information
- Patent Number: US6606624B1
Application Number: US1999467447A
Inventor: Goldberg, Kenneth Y.
Priority Date: 13 Aug 1999
Priority Number: US6606624B1
Application Date: 20 Dec 1999
Publication Date: 12 Aug 2003
IPC Current: G06F001700 | G06F001730 | G06N000500
US Class: 001001 | 707006 | 706045 | 707999006 | 707E17109
Assignee Applicant: The Regents of the University of California
Title: Apparatus and method for recommending to an individual selective information contained within a computer network
Usefulness: Apparatus and method for recommending to an individual selective information contained within a computer network
Summary: For recommending selective information such as jokes, books, movies, stocks, toys, gifts, records, music and advertisements, to client terminals such as standard computer through Internet.
Novelty: Internet based information provision method e.g. for books, involves converting multi-dimensional preference data of users into lower dimensional data using principal component analysis during off-line phase
- Industry
- ICT/Telecom
- Sub Category
- IT System
- Application No.
- 6606624
- Others
-
Tech ID/UC Case
16886/2000-010-0
Related Cases
2000-010-0
- *Abstract
-
The networked world contains a vast amount of data. Visitors face the arduous task of retrieving information that matches their preferences. The term "collaborative filtering" describes techniques that use the known preferences of a group of users to predict the unknown preferences of a new user, so that recommendations for the new user are based on these predictions. Users collaborate in the sense that each rating improves the performance of the overall system. The fundamental assumption is that if users A and B rate a number of items similarly, they share similar tastes and will rate other items similarly.
Eigentaste is a collaborative filtering algorithm that uses universal queries to elicit real-valued user ratings on a common set of items, then applies principal component analysis (PCA) to the resulting dense subset of the ratings matrix. PCA facilitates dimensionality reduction of offline clustering of users and rapid computation of recommendations.
Processing time is a constant using Eigentaste, independent of the number of users. For a large database of users, therefore, it is faster than standard nearest-neighbor techniques.
- *IP Issue Date
- Aug 12, 2003
- *Principal Investigator
-
Name: Kenneth Yigael Goldberg
Department:
- Country/Region
- USA
