Recommender systems have emerged over the last several years as an important area of research spanning the boundaries of such diverse set of disciplines as data mining, machine learning, information retrieval, human-computer interaction, marketing and operations research. Interest in recommender systems was further enhanced when Netflix announced its $1,000,000 prize competition in October 2006 that attracted over 20,000 participants from 167 different countries. One of the sub-fields of recommender systems that benefited very significantly from the Netflix Prize competition is the area of large-scale recommender systems, which deals with scaling recommendation methods to large datasets. Many Netflix competitors came to realize that some of the well-known recommendation algorithms would not scale well to the Netflix dataset. In addition, some of the most popular and well-regarded methods would perform poorly on the Netflix dataset — maybe because the asymptotic performance of these methods is quite different from their performance on smaller datasets.
