融合协同过滤的神经Bandits推荐算法  

Neural Bandits Recommendation AlgorithmBased on Collaborative Filtering

作　　者：张婷婷 欧阳丹彤[1,2] 孙成林[3] 白洪涛[1,2] ZHANG Tingting;OUYANG Dantong;SUN Chenglin;BAI Hongtao(College of Computer Science and Technology,Jilin University,Changchun 130012,China;Key Laboratory of Symbolic Computation and Knowledge Engineer of Ministry of Education,Jilin University,Changchun 130012,China;Department of Endocrinology and Metabolism,First Hospital of Jilin University,Changchun 130021,China)

机构地区：[1]吉林大学计算机科学与技术学院,长春130012 [2]吉林大学符号计算与知识工程教育部重点实验室,长春130012 [3]吉林大学白求恩第一医院内分泌与代谢科,长春130021

出　　处：《吉林大学学报（理学版）》2024年第1期92-99,共8页Journal of Jilin University:Science Edition

基　　金：吉林省自然科学基金(批准号:20210101181JC)。

摘　　要：针对数据稀疏性和“冷启动”对协同过滤的限制以及现有的协同多臂老虎机算法不适用于非线性奖励函数的问题,提出一种融合协同过滤的神经Ba ndits推荐算法COEENet.首先,采用双神经网络结构学习预期奖励及潜在增益;其次,考虑邻居协同作用;最后,构造决策器进行最终决策.实验结果表明,该方法在累积遗憾上优于4种基线算法,推荐效果较好.Aiming at the problems of the limitations of data sparsity and“cold start”on collaborative filtering and the inapplicabilit y of the existing collaborative multi-armed Bandit algorithm to nonlinear reward functions,we proposed a neural Bandit recommendation algorithm COEENet,which combined collabo rative filtering.Firstly,it adopted a dual neural network structure to learn expected rewards and potential gains.Secondly,we considered the collaborative effect of neighbors.Finally,a decision-maker was constructed to make the final decision.The experimental results show that the proposed method is superior to the four baseline algorithms in cumulative regret,and has a go od recommendation effect.

关 键 词：协同过滤 多臂老虎机算法 推荐系统 冷启动 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]