期刊名称：Bulletin of the Technical Committee on Data Engineering
出版年度：2009
卷号：32
期号：04
出版社：IEEE Computer Society
摘要：Today’s recommendation systems have evolved far beyond the initial approaches from more than a decade
ago, and have seen a great deal of commercial success (c.f., Amazon and Netflix). However, they are still far
from perfect, and face tremendous challenges in increasing the overall utility of the recommendations. These
challenges are present in all stages of the recommendation process. They begin with the cold start problem:
given a new user how do we recommend items to the user without forcing her to give feedback on a starter set of
items. Similarly, given a new item introduced into the system, how do we know when (and to whom) we should
recommend it.
