理性背叛的推荐合作激励机制  

Recommendation Incentive Mechanisms with Rational Defection for Promoting Cooperation

作　　者：金星[1] 李明楚[1] 孙晓梅 郭成[1] JIN Xing;LI Ming-chu;SUN Xiao-mei;GUO Cheng(School of Software,Dalian University of Technology.Dalian 116620.China)

机构地区：[1]大连理工大学软件学院,辽宁大连116620

出　　处：《小型微型计算机系统》2020年第2期244-251,共8页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61572095,61877007)资助.

摘　　要：为了促进协作系统中用户的合作行为,激励机制得到了广泛的使用.然而,现有的激励机制往往存在无条件合作策略占优互惠策略的现象,进而抑制了合作的涌现.为了解决这一问题,本文在推荐激励模型上进一步考虑了用户的理性背叛行为.以演化博弈为框架,研究了理性背叛机制在全局平均学习和当前最优学习两种模式下的策略演化特性.结合实际场景,本文还研究了在非完美推荐下理性背叛机制的鲁棒性问题,并且基于余弦相似度提出了一种策略识别方案.最后,通过大量的数值实验与仿真实验,验证了理性背叛机制的理论特性,也展示了该机制在促进合作方面的有效性能.To promote cooperation in collaborative systems,incentive mechanisms have been widely used. However,in existing incentive mechanisms,it is found that the unconditional cooperation strategy always dominates the reciprocity strategy,which inhibiting the emergence of cooperation. To address this issue,based on the recommendation incentive model,we further consider the defection behaviors of rational users’. And then,by using the evolutionary game theory,we study the population dynamics of the proposed rational defection model under two different learning mechanisms: GMLM( Global Mean Learning Mechanism) and CBLM( Current Best Learning Mechanism). In order to incorporate practical application,firstly,we study the robustness of our model under the scenario of imperfect recommendation,and then,based on the cosine similarity we propose a method to distinguish between strategies. At last,extensive numerical and simulation experiments are conducted to validate the theoretical properties of our model and illustrate the sufficiency of our model in promoting cooperation.

关 键 词：合作 激励机制 理性背叛 演化博弈 协作系统 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]