结合流形结构与柔性嵌入的多标签特征选择

将线性回归模型与流形结构相结合,构成了弱线性多标签特征选择的联合框架。首先,用最小二乘损失函数来学习回归系数矩阵;其次,通过标签流形结构来学习数据特征的权重矩阵;再次,用L2,1-范数来约束回归系数矩阵和特征权重矩阵,这样既能引导稀疏性,又有利于特征选择。此外,设计并证明了具有收敛性的迭代更新算法来解决上述提出的问题。最后,所提出的方法在多个经典多标签数据集上进行了验证,实验结果表明了所提算法的有效性。     

基于强化局部特征的3D点云分类与分割网络

在点云的处理过程中，许多深度学习网络未能充分考虑局部点之间的复杂关系，导致大量空间几何信息丢失。针对该问题，提出了一个强化局部特征的网络，用于点云的目标分类和语义分割。该网络通过设计编码单元对点的多方向信息进行编码；通过注意力机制学习采样分组后形成局部点云的特征。同时提出了一种新的多维损失函数，结合使用交叉熵损失函数与中心损失函数作用于分类任务。在数据集ModelNet40和ScanNet上进行了大量实验，结果表明：该网络在点云的目标分类和语义分割任务上表现出较好的性能。     

双向认知网络物理层安全策略建模及仿真研究

为提高移动通信系统的性能,构建了一个存在主网络干扰和窃听者的双向认知中继网络的物理层安全传输模型,设计了一种中继选择和功率分配的联合优化策略,用于对抗窃听者,保护收发节点的信息传输。通过仿真分析得出,此非凸优化问题限制条件较多,难以保证粒子群优化算法随机生成的初始解的可行性,从而造成求解困难。提出了一种基于可变网格优化和粒子群优化算法的混合优化算法。仿真结果表明此算法提高了次级网络的保密速率,提升了次级网络的安全性能。     

基于元学习的航空电子设备特征选择算法推荐方法

为了对航空电子设备的测试数据进行有效约简, 去除冗余信息和不相关特征, 基于机器学习领域现有的特征选择算法, 提出了一种元学习框架下的航空电子设备特征选择算法推荐方法。所提方法旨在根据不同航空电子设备测试数据所蕴含的信息, 推荐合适的特征选择算法。首先, 分析了数据集特征的描述方法。然后, 介绍了采用综合度量指数的算法性能评价方法。最后, 给出了特征选择算法推荐方法的框架。使用42个航空电子设备的测试数据和13个过滤型特征选择算法建立了元数据库, 采用留一法进行交叉验证, 推荐命中率达到了90%以上, 推荐性能比例达到97%以上。     

基于新型空间注意力机制和迁移学习的垃圾图像分类算法

随着我国各级政府大力推动垃圾强制分类,分类回收各环节中实现标准化、自动化的垃圾分类识别需要适合云端部署的高准确率、低延时要求的细粒度图像分类模型.本文发挥深度迁移学习的优势建立了一套端到端的迁移学习网络架构GANet (garbage neural network);针对垃圾分类中类别易混淆、背景干扰等挑战,提出一种新型的像素级空间注意力机制PSATT (pixel-level spatial attention).为克服类别多和样本不平衡挑战,提出使用标签平滑正则化损失函数;为改善收敛速度以及模型稳定性与泛化性,提出了阶梯形OneCycle学习率控制方法,并给出了结合Rectified Adam (RAdam)优化方法和权重平滑处理技术的组合使用策略.实验使用了"华为云人工智能大赛.垃圾分类挑战杯"提供的按照深圳市垃圾分类标准标注的训练数据,验证了GANet在垃圾分类问题中的显著效果,获得了全国二等奖(第2名);同时,提出的PSATT机制优于对比方法,且在不同主干网络架构上均得到了提升,具有良好的通用性.本文提出的GANet架构、PSATT机制和训练策略不仅具有重要的工程参考价值,也具有较好的学术价值.     

基于特征选取与树状Parzen估计的入侵检测

针对目前网络空间安全形势快速变化带来的新风险和新挑战, 提出一种基于相关性分析的特征选取和树状Parzen估计优化的入侵检测方法。首先, 通过基于相关性分析的数据特征选取方法对数据维度进行压缩。其次, 对原始数据集进行特征筛选, 生成新的特征子集。最终, 使用序列模型优化算法中的树状Parzen估计算法对随机森林算法进行模型优化。实验结果表明, 相比其他应用机器学习算法的入侵检测方法, 所提方法在提升综合性能的同时拥有更高的检测效率, 有效地提升了入侵检测技术的实用性。     

Space use by two arboreal rodent species in a Neotropical cloud forest

Studying animal space use patterns can help increase our understating of ecological processes such as competition and community dynamics. To quantify space and habitat use in an isolated and patchy cloud forest community in Mexico, we evaluate the vertical stratification, home range and habitat selection of two arboreal rodents: Habromys schmidlyi and Reithrodontomys microdon. Using live-traps at ground level and different forest strata, we radio-equipped nine individuals of H. schmidlyi and seven of R. microdon, and evaluated fine-scale space use and broad-scale habitat selection between cloud forest and oak forest. We found an average home range of 0.24 ha for R. microdon males and 0.72 ha for females, with a preference for higher canopy in the cloud forest. For H. schmidlyi the home range was 0.83 ha for males and 0.29 ha for females, with a preference for the understory level in the cloud forest. Home range is three-dimensional for these rodents, so we estimate that on average, individuals of both species used eight trees in the time they were tracked. We characterised the vegetation at the trap sites, and used recursive partitioning to relate the presence of different plants with the probability of finding these two species and Peromyscus aztecus, a third rodent species also present in the area and considered in our analysis of habitat use. The highest probability of finding R. microdon (96%) was related to the presence of Brachythecium occidentale and Renauldia mexicana, while H. schmidlyi (95%) was found in close proximity to Fabronia ciliaris and Everniastrum. We highlight the importance of arboreal trapping in biodiversity assessments, and the role of arboreal rodents in maintaining tropical forest ecosystems. We suggest that these rodent species could avoid or reduce competition by using the vertical strata differentially, and that H. schmidlyi and R. microdon can be biological indicators for cloud forest management and conservation.     

基于变量选择-神经网络模型的复杂路网短时交通流预测

摘要： 针对传统交通流预测模型正在由单断面历史数据处理向多断面、多时刻历史数据处理转变,但在考虑各断面间的影响时,多变的交通状况往往会使预测模型复杂化的问题,引入一种多元线性回归最小绝对收缩和选择算子方法(Lasso),并利用其优秀的变量选择能力,在复杂路网多断面中选出相关性较高的断面；结合神经网络(NN)的非线性特性,提出了Lasso NN组合模型.结果表明：Lasso NN模型在路网交叉口对未来15 min交通流数据预测的误差率低于9.2%；在非交叉口的误差率低于6.7%,总体优于各自单独使用得出的结果.     

选材引起测量金属丝杨氏模量的误差分析

杨氏模量的测量是大学物理中重要的力学实验之一。我们对不同直径、不同长度的钢丝进行了杨氏模量测定,分析了选材对实验结果的影响。研究结果表明,在仪器允许的测量范围内,直径越小、长度越长的材料,其杨氏模量的相对合成标准不确定度越小。     

Dynamic Model Averaging and CPI Inflation Forecasts: A Comparison between the Euro Area and the United States

The paper forecasts consumer price inflation in the euro area (EA) and in the USA between 1980:Q1 and 2012:Q4 based on a large set of predictors, with dynamic model averaging (DMA) and dynamic model selection (DMS). DMA/DMS allows not solely for coefficients to change over time, but also for changes in the entire forecasting model over time. DMA/DMS provides on average the best inflation forecasts with regard to alternative approaches (such as the random walk). DMS outperforms DMA. These results are robust for different sample periods and for various forecast horizons. The paper highlights common features between the USA and the EA. First, two groups of predictors forecast inflation: temporary fundamentals that have a frequent impact on inflation but only for short time periods; and persistent fundamentals whose switches are less frequent over time. Second, the importance of some variables (particularly international food commodity prices, house prices and oil prices) as predictors for consumer price index inflation increases when such variables experience large shocks. The paper also shows that significant differences prevail in the forecasting models between the USA and the EA. Such differences can be explained by the structure of these respective economies. Copyright © 2015 John Wiley & Sons, Ltd.