基于自适应协同聚类的支持向量预选算法

为提高支持向量机在大规模数据集上的训练效率,提出一种基于自适应协同聚类的支持向量预选算法。该方法通过对两类样本进行自适应协同聚类,寻找少量具有协同关系的类中心对,替代支持向量进行训练,从而减少参训样本数量。其中,中心对数量由算法自动确定。与其他支持向量预选算法的对比实验结果表明,自适应协同聚类算法能够在不影响分类性能的情况下,有效提高训练速度,是一种行之有效的快速支持向量预选算法。     

无线传感器网络的自适应分簇组网算法研究

针对经典算法LEACH和HEED的不足,提出了一种能自适应分簇组网的优化算法。构建了节点信息权重模型,并借鉴邻节点信息交换的思想,在成簇过程中与探测范围内的节点交换权重信息,自适应完成分布式网络的簇首选举,并根据最小距离原则成簇。理论分析和仿真实验表明,该算法比LEACH和HEED算法选取的簇首及形成的簇结构更加合理,同时更有效地降低与均衡了网络的能耗,提高了传感器网络的生命周期。

Abstract：

A self-adaptive and optimized clustering algorithm was put forward according to the shortage of LEACH and HEED. The Heavy-weight model about nodes＇ messages was created,and the idea that neighbor nodes exchanged messages each other was used for reference. The nodes which were able to communicate with each other exchanged the Heavy-weight during making clusters,elect self-adaptively the cluster head in distributing networks,and made some clusters based on the minimum distance principle. The theoretic analysis and simulation results prove that the elected cluster head and cluster structure are more reasonable,the energy expenditure in networks is less,the longevity of networks is longer by the optimized algorithm compared to LEACH and HEED.     

衰落信道WSN网络基于投票机制的决策融合

针对衰落信道网络中传感节点难以准确获取检测信息的问题,构建了并行结构分布检测的系统模型,提出了一种基于投票机制的决策融合算法。通过邻居节点间的信息交互,各传感节点获取了通讯半径内邻居节点的判决,并根据多票优先的原则重新调整自身决策,提高了检测的准确性。理论分析和仿真实验表明,该算法比传统的EGC融合规则具有更高的检测概率和稳定性,适合中等规模无线传感器网络。