[1]吴 杰,姜宜鑫*,韩国敬,等.一种基于变分推断的可评判推荐算法[J].计算机技术与发展,2023,33(10):150-156.[doi:10. 3969 / j. issn. 1673-629X. 2023. 10. 023]
　WU Jie,JIANG Yi-xin*,HAN Guo-jing,et al.A Critiquing Recommendation Algorithm Based on Variational Inference[J].,2023,33(10):150-156.[doi:10. 3969 / j. issn. 1673-629X. 2023. 10. 023]

点击复制

一种基于变分推断的可评判推荐算法()

分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:

33

期数:

2023年10期

页码:

150-156

栏目:

人工智能

出版日期:

2023-10-10

文章信息/Info

Title:

A Critiquing Recommendation Algorithm Based on Variational Inference

文章编号:

1673-629X(2023)10-0150-07

作者:

吴 杰¹ ; 姜宜鑫^1* ; 韩国敬² ; 马 驰³

1. 辽宁科技大学 计算机与软件工程学院,辽宁 鞍山 114051;

2. 鞍山市气象局,辽宁 鞍山 114004;

3. 惠州学院 计算机科学与工程学院,广东 惠州 516007

Author(s):

WU Jie1 ; JIANG Yi-xin1* ; HAN Guo-jing2 ; MA Chi3

1. School of Computer and Software Engineering,University of Science and Technology Liaoning,Anshan 114051,China;

2. Anshan Meteorological Bureau,Anshan 114004,China;3. School of Computer Science and Engineering,Huizhou University,Huizhou 516007,China

关键词:

推荐算法; 变分推断; 神经协同过滤; 可评判; 可解释性

Keywords:

recommendation algorithm; variational inference; neural collaborative filtering; critiquing; interpretability

分类号:

TP301

DOI:

10. 3969 / j. issn. 1673-629X. 2023. 10. 023

摘要:

随着互联网时代的不断发展,互联网上的信息量不断增多,“ 信息过载” 等相关问题愈发严重,从而导致用户很难快速地获取到有用的信息,因此推荐系统应运而生。推荐系

统可以预测用户的需求并推荐给用户其最可能喜欢的内容,来缓解人们从海量信息中做出选择的烦恼。 推荐算法是推荐系统的核心,它完全可以决定一个推荐系统的性能。 推荐准确度及可解释性是推荐算法目前面临的两大难题。 可评判推荐算法是对话推荐算法的一种,在预测出项目的同时,也及时给出推荐项目的理由,并且为用户提

供一个重新推荐的机会,用户通过对解释项进行评判来使推荐系统重新预测出商品,可有效解决上述两个问题。 该文首先基于变分推断与神经协同过滤相结合的思想,对算法和模型进行了形式化的定义和理论推导,并且从概率的角度出发使用贝叶斯神经网络实现了该模型。 通过与其他可评判推荐算法进行实验对比,证实了该模型的许多推荐指标已经达到了目前最先进的水平。

Abstract:

With the continuous development of the Internet era,the amount of information on the Internet is increasing,and the relatedproblems such as " information overload" are becoming more and more serious,which makes it difficult for users to quickly obtain usefulinformation. Therefore, the recommendation system comes into being. Recommendation systems can predict users’ needs andrecommend what they are most likely to like,relieving people of the annoyance of choosing from a vast amount of information. Recommendation algorithm is the core of recommendation system,which can determine the performance of a recommendation system. Recommendation accuracy and interpretability are two major problems facing recommendation algorithms at present. Evaluable recommendationalgorithm is a kind?

of dialogue recommendation algorithm. While predicting the items,it also gives reasons for the recommended items intime,and provides users with an opportunity to?

re- recommend. Users can make the recommendation system re - predict the goods byjudging the explanatory items,which can effectively solve the above two problems. Based on the idea of combining variational inferenceand neural collaborative filtering,we firstly formalize the definition and theoretical derivation of the algorithm and model,and then realizethe model from the perspective of probability by using Bayesian neural network. Through the experimental comparison with otherevaluable recommendation algorithms,it is confirmed that many recommendation indexes of this model have reached the most advancedlevel.

常用功能