链路预测是利用深度学习技术分析网络数据,挖掘网络中潜在的节点关系,通常应用于网络安全、信息挖掘等领域。通过预测网络中节点间的链路,可以识别社交工程攻击、欺诈行为和隐私泄露风险。但移动社交网络的拓扑结构随时间变化,链路稀疏...链路预测是利用深度学习技术分析网络数据,挖掘网络中潜在的节点关系,通常应用于网络安全、信息挖掘等领域。通过预测网络中节点间的链路,可以识别社交工程攻击、欺诈行为和隐私泄露风险。但移动社交网络的拓扑结构随时间变化,链路稀疏,影响预测准确性。为了解决移动社交网络中链路预测的强稀疏性问题,提出基于深度学习的预测方法,即面向强稀疏性移动社交网络的链路预测深度学习方法(deep learning-based method for mobile social networks with strong sparsity for link prediction,DLMSS-LP)。该方法综合运用了图自编码器(graph auto-encoder,GAE)、特征矩阵聚合技术以及多层长短期记忆网络(long short-term memory,LSTM),旨在降低了模型的学习成本,更有效地处理高维和非线性的网络结构,并且捕捉移动社交网络中的时序动态变化,进而增强模型对现有链路生成可能性的预测能力。对比其他方法在AUC(area under curve)和ER(error rate)指标上有明显提升,体现了模型对不确定链路预测的高准确率和强鲁棒性。展开更多
在A d Hoc网络中,节点的频繁移动导致链路经常失效,AODV路由协议对失效链路反应速度过慢,使网络中报文丢失率增加以及端到端平均传递时延增长。为了解决这个问题,文章提出了一种路由切换的算法。使活动路由中的每个节点收到数据报文时...在A d Hoc网络中,节点的频繁移动导致链路经常失效,AODV路由协议对失效链路反应速度过慢,使网络中报文丢失率增加以及端到端平均传递时延增长。为了解决这个问题,文章提出了一种路由切换的算法。使活动路由中的每个节点收到数据报文时估算链路的状态,如果发现正在使用的链路即将失效,则节点在链路失效前将相关路由信息切换到合适的节点上。通过ns-2对增加切换算法的AODV协议进行仿真,结果表明,在节点移动的情况下,改进后的算法明显提高了AODV协议的报文投递率,降低了端到端平均传递时延,而路由开销仅有少量的增加。展开更多
In an abnormal high-temperature fire environment,a structure with mechanical-thermal weak-link can be used to predict the permanent failure before the failure of some strong-links,such as explosive initiator,thus to p...In an abnormal high-temperature fire environment,a structure with mechanical-thermal weak-link can be used to predict the permanent failure before the failure of some strong-links,such as explosive initiator,thus to protect the latter. An effective design can be defined as that the weak-link fails before the failure of the strong-link,and the system is safe; while an unsuccessful design means that the weak-link fails after the failure of the strong-link,and the system loses in safety. The probability of safety failure exists due to the uncertain failure temperatures of the weak-link and strong-link. In order to obtain the probability of safety failure,a statistical method was used to deal with the uncertainty of the failure temperatures. The integral method and stochastic simulation method were used in calculations. Finally,a sample was given to verify the consistence of the results given by two methods.展开更多
文摘链路预测是利用深度学习技术分析网络数据,挖掘网络中潜在的节点关系,通常应用于网络安全、信息挖掘等领域。通过预测网络中节点间的链路,可以识别社交工程攻击、欺诈行为和隐私泄露风险。但移动社交网络的拓扑结构随时间变化,链路稀疏,影响预测准确性。为了解决移动社交网络中链路预测的强稀疏性问题,提出基于深度学习的预测方法,即面向强稀疏性移动社交网络的链路预测深度学习方法(deep learning-based method for mobile social networks with strong sparsity for link prediction,DLMSS-LP)。该方法综合运用了图自编码器(graph auto-encoder,GAE)、特征矩阵聚合技术以及多层长短期记忆网络(long short-term memory,LSTM),旨在降低了模型的学习成本,更有效地处理高维和非线性的网络结构,并且捕捉移动社交网络中的时序动态变化,进而增强模型对现有链路生成可能性的预测能力。对比其他方法在AUC(area under curve)和ER(error rate)指标上有明显提升,体现了模型对不确定链路预测的高准确率和强鲁棒性。
文摘在A d Hoc网络中,节点的频繁移动导致链路经常失效,AODV路由协议对失效链路反应速度过慢,使网络中报文丢失率增加以及端到端平均传递时延增长。为了解决这个问题,文章提出了一种路由切换的算法。使活动路由中的每个节点收到数据报文时估算链路的状态,如果发现正在使用的链路即将失效,则节点在链路失效前将相关路由信息切换到合适的节点上。通过ns-2对增加切换算法的AODV协议进行仿真,结果表明,在节点移动的情况下,改进后的算法明显提高了AODV协议的报文投递率,降低了端到端平均传递时延,而路由开销仅有少量的增加。
基金Sponsored by Science and technology development fund of China academy of engineering physics( 2011A0203010)
文摘In an abnormal high-temperature fire environment,a structure with mechanical-thermal weak-link can be used to predict the permanent failure before the failure of some strong-links,such as explosive initiator,thus to protect the latter. An effective design can be defined as that the weak-link fails before the failure of the strong-link,and the system is safe; while an unsuccessful design means that the weak-link fails after the failure of the strong-link,and the system loses in safety. The probability of safety failure exists due to the uncertain failure temperatures of the weak-link and strong-link. In order to obtain the probability of safety failure,a statistical method was used to deal with the uncertainty of the failure temperatures. The integral method and stochastic simulation method were used in calculations. Finally,a sample was given to verify the consistence of the results given by two methods.
