网络结构发现可识别网络多类型聚类模式,但其准确率有待提升.批量主动学习选择质量高的节点集合构造先验,可提升无监督网络结构发现的性能.面向属性网络分类的主动学习BMAL(Batch Mode Active Learning)只考虑链接信息实现网络节点选择...网络结构发现可识别网络多类型聚类模式,但其准确率有待提升.批量主动学习选择质量高的节点集合构造先验,可提升无监督网络结构发现的性能.面向属性网络分类的主动学习BMAL(Batch Mode Active Learning)只考虑链接信息实现网络节点选择,但不能有效选择使模型性能提升至最优的节点集合,且依赖初始人工标注及参数.提出一个新的批量主动学习算法,利用目标函数的子模性迭代选择最优的节点集合.该方法基于未标记节点的不确定性和非冗余影响力选择最优节点集合,不确定性依据节点及其邻居的类隶属度,影响力依据节点的非重叠中心性,两个指标的权重依据熵权法自动确定.人工和真实网络上的实验结果表明,该方法能选择使结构发现性能提升最大的节点集合.展开更多
This paper presents an overlay network topology called Virtual and Dynamic Hierarchical Architecture (VDHA) for discovering Grid services with high performance. Service discovery based on VDHA has scalable, autonomous...This paper presents an overlay network topology called Virtual and Dynamic Hierarchical Architecture (VDHA) for discovering Grid services with high performance. Service discovery based on VDHA has scalable, autonomous, efficient, reliable and quick responsive. We propose two service discovery algorithms. Full Search Query and Discovery Protocol (FSQDP) discovers the nodes that match the request message from all N nodes, which has time complexity O(logN), space complexity O(nvg) (nvg being node numbers of each virtual group), and message-cost O(N), and Domain-Specific Query and Discovery Protocol (DSQDP) searches nodes in only specific domains with time complexity O(nvg), space complexity O(nvg), and message-cost O(nvg). In this paper, we also describe VDHA, its formal definition, and Grid Group Management Protocol.展开更多
文摘网络结构发现可识别网络多类型聚类模式,但其准确率有待提升.批量主动学习选择质量高的节点集合构造先验,可提升无监督网络结构发现的性能.面向属性网络分类的主动学习BMAL(Batch Mode Active Learning)只考虑链接信息实现网络节点选择,但不能有效选择使模型性能提升至最优的节点集合,且依赖初始人工标注及参数.提出一个新的批量主动学习算法,利用目标函数的子模性迭代选择最优的节点集合.该方法基于未标记节点的不确定性和非冗余影响力选择最优节点集合,不确定性依据节点及其邻居的类隶属度,影响力依据节点的非重叠中心性,两个指标的权重依据熵权法自动确定.人工和真实网络上的实验结果表明,该方法能选择使结构发现性能提升最大的节点集合.
文摘This paper presents an overlay network topology called Virtual and Dynamic Hierarchical Architecture (VDHA) for discovering Grid services with high performance. Service discovery based on VDHA has scalable, autonomous, efficient, reliable and quick responsive. We propose two service discovery algorithms. Full Search Query and Discovery Protocol (FSQDP) discovers the nodes that match the request message from all N nodes, which has time complexity O(logN), space complexity O(nvg) (nvg being node numbers of each virtual group), and message-cost O(N), and Domain-Specific Query and Discovery Protocol (DSQDP) searches nodes in only specific domains with time complexity O(nvg), space complexity O(nvg), and message-cost O(nvg). In this paper, we also describe VDHA, its formal definition, and Grid Group Management Protocol.
