Bidirectional Label Propagation over Graphs

Graph-Based label propagation algorithms are popular in the state-of-the-art semi-supervised learning research. The key idea underlying this algorithmic family is to enforce labeling consistency between any two examples with a positive similarity. However, negative similarities or dissimilarities are equivalently valuable in practice. To this end, we simultaneously leverage similarities and dissimilarities in our proposed semi-supervised learning algorithm which we term Bidirectional Label Propagation (BLP). Different from previous label propagation mechanisms that proceed along a single direction of graph edges, the BLP algorithm can propagate labels along not only positive but also negative edge directions. By using an initial neighborhood graph and class assignment constraints inherent among the labeled examples, a set of class-specific graphs are learned, which include both positive and negative edges and thus reveal discriminative cues. Over the learned graphs, a convex propagation criterion is carried out to ensure consistent labelings along the positive edges and inconsistent labelings along the negative edges. Experimental evidence discovered in synthetic and real-world datasets validates excellent performance of the proposed BLP algorithm.     

动态环境下的主动蠕虫攻击分析

鉴于当前很少有传播模型充分考虑到P2P节点动态特征对主动蠕虫攻击的影响, 提出两个动态环境下的主动蠕虫传播模型。分析了主动蠕虫两种常见的攻击方式, 给出了相应攻击背景下的节点状态转换过程, 在综合考虑P2P节点动态特征的基础上提出了两种主动蠕虫传播模型, 并对所提出的模型进行了数值分析, 探讨动态环境下影响主动蠕虫传播速度的关键因素。实验结果表明, 通过提高P2P节点的离线率和免疫力可以有效地抑制主动蠕虫对P2P网络的攻击。     

Effect of variation in gas distribution on explosion propagation characteristics in coal mines

In order to investigate the effect of variation in the distribution of gas on explosion propagation characteristics in coal mines, experiments were carried out in two different channels with variation in gas concentration and geometry. Flame and pressure transducers were used to track the explosion front velocity. The flame speed (Sf) showed a slight downward trend while the methane concentration varied from 10% to 3% in the experimental channel. The peak overpressure (Pmax) dropped dramatically when com-pared with normal conditions. As well, the values of Pmax and Sf decreased when the methane concentration dropped from 8% to 6%. The flame speed in the channel, connected to a cylinder with a length varying from 0.5 to 2 m, was greater than that in the normal channel. The peak overpressure was also higher than that under normal conditions because of a higher flame speed and stronger pressure piling up. The values of Pmax and Sf increased with an increase in cylinder length. The research results indicate that damage caused by explosions can be reduced by decreasing the gas concentration, which should be immediately detected in roadways with large cross-sections because of the possibility of greater destruction caused by more serious explosions.     

基于近邻消息传递的自适应局部线性嵌入

针对局部线性嵌入(LLE)降维算法中邻域参数的人工设定及其全局性的缺陷,研究了聚类和降维的内在联系,提出了邻域参数的自适应选取策略,构建了一种聚类和降维的自适应局部线性嵌入(ALLE)算法,为每个样本点设计最佳的近邻搜索空间,自适应选取邻近点计算权值重建矩阵,基于类信息重新定义了重构误差函数。实验表明,新算法更能体现出数学上流形概念的局部坐标化本质,对不相关数据、冗余数据和噪声数据具有良好的鲁棒性,在实际识别问题中体现出优越的性能。     

周期阵列压痕奇异应力场与开裂（PartⅡ：压痕开裂临界载荷）

周期阵列压痕方法是研究微观摩擦学的工具之一。由于周期阵列刚性压头压在半平面弹性体的表面时,会产生周期的Ⅰ型奇异压应力场,该奇异应力场的应力集中达到一定程度时,即达到临界条件时,该奇异应力场所在的边界将开裂。压痕边界的开裂问题是一个新型的断裂力学问题,通过边界移动的能量释放率的分析给出了压痕边界开裂能量释放率数学表征,并给出边界开裂的临界载荷和临界开裂角。     

基于用户行为和网络拓扑的Email蠕虫传播

通过分析用户行为规律和电子邮件网络拓扑特征,提出了能够准确描述Email蠕虫传播特性的仿真算法。首先,基于Enron Email数据集构建了能够表征Email网络特点的仿真环境,实验结果验证了仿真算法能够准确地体现Email蠕虫传播特性。然后从理论和实验两方面分析验证了关键节点被感染后会加速蠕虫的传播的结论,并讨论了不同防护措施对蠕虫传播的抑制效果。最后建立了Email蠕虫传播数学解析模型(Topo-SIS),与仿真结果的比较表明,该模型能够较好地预测Email蠕虫的传播规模。     

震后裂缝危岩体诊断技术研究

震后山体结构破碎,崩塌频发,严重的制约着灾后重建工作。为此,从震后岩体实际结构状况出发,在分析诱发岩体崩塌的破坏性外力类型的基础上,借助断裂力学理论,通过合理建模,阐明了震后岩体崩塌的机理,给出了崩塌发生的条件,为震区可预知余震强度区的危岩体诊断和排查提供了技术支持。最后通过算例及分析表明：余震、降雨及其相互耦合作用是震后崩塌的根本动力;在特定外力下区内岩体裂缝发生扩展存在最小临界深度;外力耦合条件下裂缝失稳临界深度最小     

基于BP神经网络的集装箱箱号识别

提出一种用BP神经网络实现集装箱箱号识别的方法。提出的图像预处理方法对管柱,底面凹凸,脏污物、光线的强弱、倾斜、缺损或模糊有较强的适应性。在BP网络分类器实现时,考虑集装箱箱号的组成特点,采用并行处理方法。在BP算法中,引入动态因子,动态调整学习步长。实践证明利用这种方法来识别装箱箱号,速度快,识别率高,具有较高的实用价值。     

人工神经网络在改善报警系统性能中的应用

介绍了一种利用BP人工神经网络降低报警系统误报警率的有效方法。以STC89C52RC处理器为平台,在嵌入式系统中应用BP网络先对报警信息进行分析和预处理,再控制报警系统动作。该方法大大降低了报警系统因自然因素导致的误报警概率。采集了100个典型数据对BP网络进行训练,得到了适合应用目标的网络模型,现场实测误报警率几乎为0。该方法的硬件价格便宜,电路简单,是一种降低报警系统误报警率的有效方法。     

一种三维复杂介质波前时间的高精度数值算法

基于水平集的波前扩展算法,如FMM(Fast Marching Method)、GMM(Group Marching Method),作为一类计算复杂介质波前时间的有效方法而被广泛使用.该类算法都是基于程函方程的有限差分格式来计算波传播时间,在介质离散单元尺寸较大的情况下,计算精度较低.为提高波前时间的计算精度,在一个长方体单元内,将任意点的波传播时间用已知节点上波前时间的插值函数表示,然后根据Fermat原理确定未知节点上的波前时间,再结合高效率的GMM算法,形成了一种计算三维复杂介质波前时间的有效算法.数值模拟实验表明,与原GMM算法相比,该算法大大提高了波前时间的计算精度,同时具有很强的稳定性和适应性.