非独立同分布推荐系统:推荐范式转换的综述和框架

发布时间：2017-12-27 13:02

本文关键词：非独立同分布推荐系统:推荐范式转换的综述和框架　出处：《Engineering》2016年02期 　论文类型：期刊论文

【摘要】：虽然推荐系统在我们的生活、学习、工作和娱乐中扮演着越来越重要的角色,但是很多时候我们收到的推荐都是不相关的、重复的,或者包含不感兴趣的产品和服务。这些差的推荐系统产生的原因来源于一个本征假设:传统的理论和推荐系统认为用户和物品是独立同分布的(IID)。另一个明显的现象是,虽然投入了很多的精力模拟用户或者物品的特殊属性,但用户和物品的总体属性及它们之间的非独立同分布性(non-IID)被忽略了。本文先讨论了推荐系统的非独立同分布性,紧接着介绍了非独立同分布性原理,目的是从耦合和异构性的角度来深入阐述传统的推荐系统的固有本质。这种非独立同分布推荐系统引起了传统推荐系统范式的转化——从独立同分布向非独立同分布进行转化,希望能够形成高效的、相关性高的、个人订制和可操作的推荐系统。这种系统创造了令人兴奋的能够解决包含冷启动、以稀疏数据为基础、跨域、基于群组信息和欺诈攻击等各种复杂情况的新的研究方向和解决方案。
[Abstract]:Although recommender systems play an increasingly important role in our life, study, work and entertainment, many times, the recommendations we receive are irrelevant, repetitive or contain products and services that are not interested. The reasons for these poor recommendation systems come from an eigenhypothesis: traditional theories and recommender systems believe that users and items are independently and identically distributed (IID). Another obvious phenomenon is that although a lot of effort has been invested in simulating the special attributes of users or articles, the overall attributes of users and items and the non independent and identically distributed (non-IID) between them have been ignored. This paper first discusses the independent and identically distributed nature of recommender systems, and then introduces the principle of non identity and identically distributed. The purpose is to further explain the inherent nature of traditional recommender systems from the perspective of coupling and heterogeneity. This kind of independent and identically distributed recommender system has caused the transformation of traditional recommender system paradigm -- transforming from independent distribution to independent distribution, hoping to form an efficient, highly correlated, personal customized and operable recommender system. This system has created an exciting new research direction and solution to solve the complex situations including cold start, sparse data based, cross domain, group based information and fraud attacks.
【作者单位】： Advanced
【分类号】：TP391.3
【正文快照】： 1.IntroductionRecommendation[1,2]is a major application of big data[3,4].It plays an increasingly important role in both core business andnew economy,particularly when it involves social media,mobileservices,online business,and study and living.In recent

【相似文献】