A more efficient Collaborative Filtering method

Speaker: TAM Ming Wai

Abstract

Collaborative Filtering is simulating the process of looking for opinions from similar-taste friends. An important part of CF is finding the top-k most similar users. The nature of this top-k search differs from the conventional top-k problem. This talk will introduce the basic ideas of inverted-file-based CF, as well as the alternative methods of item-based CF and slope-one CF. A new, hybrid method will be presented, which combines sequential access to an inverted file and random access.