针对移动边缘计算中用户移动性导致服务器间负载分布不均,用户服务质量(Quality of Service,QoS)下降的问题,提出了一种移动性感知下的分布式任务迁移方案。首先,以优化网络中性能最差的用户QoS为目标,建立了一个长期极大极小化公平性问...针对移动边缘计算中用户移动性导致服务器间负载分布不均,用户服务质量(Quality of Service,QoS)下降的问题,提出了一种移动性感知下的分布式任务迁移方案。首先,以优化网络中性能最差的用户QoS为目标,建立了一个长期极大极小化公平性问题(Max Min Fairness,MMF),利用李雅普诺夫(Lyapunov)优化将原问题转化解耦。然后,将其建模为去中心化部分可观测马尔可夫决策过程(Decentralized Partially Observable Markov Decision Process,Dec-POMDP),提出一种基于多智能体柔性演员-评论家(Soft Actor-Critic,SAC)的分布式任务迁移算法,将奖励函数解耦为节点奖励和用户个体奖励,分别基于节点负载均衡度和用户QoS施加奖励。仿真结果表明,相比于现有任务迁移方案,所提算法能够在保证用户QoS的前提下降低任务迁移率,保证系统负载均衡。展开更多
随着互联网的发展,信息爆炸的问题日益突出,开发者很难准确地找到自己需要的服务。服务推荐技术可以帮助开发者过滤信息,提供个性化的推荐,节省开发者的时间和精力。现有的Web服务推荐系统通常更注重Web服务推荐的准确性,而忽略了Web服...随着互联网的发展,信息爆炸的问题日益突出,开发者很难准确地找到自己需要的服务。服务推荐技术可以帮助开发者过滤信息,提供个性化的推荐,节省开发者的时间和精力。现有的Web服务推荐系统通常更注重Web服务推荐的准确性,而忽略了Web服务之间的兼容性。这可能导致在创建应用程序时出现服务之间不兼容的情况。为了解决上述挑战,本文提出了一种基于兼容性感知的服务推荐方法(SRCR)。首先,SRCR使用图神经网络算法来对历史记录进行深入挖掘,提取应用程序和服务的历史记录特征,从而计算其偏好。其次,SRCR通过对历史服务共同调用情况进行分析,预测候选服务和现有服务的兼容性。最后,将上述二者相融合得到最终的服务列表。在ProgrammableWeb上收集的真实数据集上进行的大量实验证明了我们所提出的SRCR方法的有效性。With the development of the Internet, the problem of information explosion has become increasingly prominent. Developers find it difficult to accurately find the services they need. Service recommendation technology can help developers filter information, provide personalized recommendations, and save developers time and energy. The existing Web service recommendation systems usually focus more on the accuracy of Web service recommendations, while ignoring the compatibility between Web services. This may lead to incompatibility between services when creating applications. To address the aforementioned challenge, this paper proposes a Service Recommendation method based on Compatibility Awareness (SRCR). Firstly, SRCR employs graph neural network algorithm to deeply mine historical records, extract historical record features of applications and services, and calculate the preferences of applications. Secondly, SRCR predicts the compatibility between candidate services and existing services by analyzing the joint invocation of historical services. Finally, the above two are combined to obtain the final service list. A large number of experiments conducted on real datasets collected on ProgrammableWeb have demonstrated the effectiveness of our proposed SRCR method.展开更多
文摘随着互联网的发展,信息爆炸的问题日益突出,开发者很难准确地找到自己需要的服务。服务推荐技术可以帮助开发者过滤信息,提供个性化的推荐,节省开发者的时间和精力。现有的Web服务推荐系统通常更注重Web服务推荐的准确性,而忽略了Web服务之间的兼容性。这可能导致在创建应用程序时出现服务之间不兼容的情况。为了解决上述挑战,本文提出了一种基于兼容性感知的服务推荐方法(SRCR)。首先,SRCR使用图神经网络算法来对历史记录进行深入挖掘,提取应用程序和服务的历史记录特征,从而计算其偏好。其次,SRCR通过对历史服务共同调用情况进行分析,预测候选服务和现有服务的兼容性。最后,将上述二者相融合得到最终的服务列表。在ProgrammableWeb上收集的真实数据集上进行的大量实验证明了我们所提出的SRCR方法的有效性。With the development of the Internet, the problem of information explosion has become increasingly prominent. Developers find it difficult to accurately find the services they need. Service recommendation technology can help developers filter information, provide personalized recommendations, and save developers time and energy. The existing Web service recommendation systems usually focus more on the accuracy of Web service recommendations, while ignoring the compatibility between Web services. This may lead to incompatibility between services when creating applications. To address the aforementioned challenge, this paper proposes a Service Recommendation method based on Compatibility Awareness (SRCR). Firstly, SRCR employs graph neural network algorithm to deeply mine historical records, extract historical record features of applications and services, and calculate the preferences of applications. Secondly, SRCR predicts the compatibility between candidate services and existing services by analyzing the joint invocation of historical services. Finally, the above two are combined to obtain the final service list. A large number of experiments conducted on real datasets collected on ProgrammableWeb have demonstrated the effectiveness of our proposed SRCR method.
