A distributed routing algorithm for data aggregation in wireless sensor networks

Considering the impact of aggregation cost on the performance of aggregation routes in wireless sensor networks, an aggregation-decision-based distributed routing algorithm for data aggregation is proposed. When source nodes arrive or leave, the algorithm can calculate the aggregation benefit according to data correlation, aggregation cost and transmission cost. Then the algorithm will adaptively make aggregation and routing decisions based on aggregation benefit. Therefore, it can jointly optimize the aggregation and transmission costs and reduce the energy consumption for data gathering. This distributed algorithm makes all the decisions only relying on the local information. Hence, the routing maintenance cost is limited. Simulation results show that the energy consumption difference between this distributed online algorithm and the previous offline one is within 17% under any network conditions. __________ Translated from Journal of Beijing University of Posts and Telecommunications, 2007, 30(1): 9–13 [译自: 北京邮电大学学报]     

面向微博短文本的社交与概念化语义扩展搜索方法

充分挖掘微博短文本的语义以实现精准搜索是一项重要任务.由于微博文本内容具有稀疏性和语义局限性的特点,使得仅通过分析字面语义来进行短文本理解和相似性匹配的传统搜索方法受到了一定的限制.因此提出了一种社交与概念化语义结合的扩展搜索方法,通过挖掘社交网络独特的社交属性如#标签#、“@”和链接信息URL,对微博短文本实现进一步的社交语义扩展.该方法将文本字面分析获取的概念词语和社交关系中潜在的关联标签信息相结合,对短文本进行2种角度下的语义特征表示,实现了基于微博短文本语义充分理解的精准搜索.在微博数据集上的对比实验表明,与已有的扩展搜索方法相比所提方法能捕捉更多的语义特征,微博搜索的性能也得到了显著的提升.     

MapReduce 环境下的并行复杂网络链路预测

为使链路预测应用于大型复杂网络,设计并实现了一种基于MapReduce计算模型的并行链路预测算法,包含了9种基于局部信息的相似性指标,在稀疏网络上的时间复杂度为O(N).首先,在公共数据集上验证了并行算法的有效性,随着抽取因子的增加,召回率升高而准确率下降.在不同类型的10个大规模复杂网络数据集上的实验结果表明,基于MapReduce计算模型的并行链路预测算法比传统算法具有更高的效率,算法的运行时间随着并行程度的增加而下降.提出并证明了AUC(area under a receiver operating characteristic curve)评价指标的上下界,实验表明,上下界的中值和实际AUC值很接近,并且AUC评价指标侧重于预测分数值是否为0而不是分数值的大小.在网络拓扑性质中,平均聚集系数对AUC值的影响最大,并且AUC值随着网络平均聚集系数的增加而提高.