Mapping landslide susceptibility at the Three Gorges Reservoir, China, using gradient boosting decision tree,random forest and information value models

This work was to generate landslide susceptibility maps for the Three Gorges Reservoir(TGR) area, China by using different machine learning models. Three advanced machine learning methods, namely, gradient boosting decision tree(GBDT), random forest(RF) and information value(InV) models, were used, and the performances were assessed and compared. In total, 202 landslides were mapped by using a series of field surveys, aerial photographs, and reviews of historical and bibliographical data. Nine causative factors were then considered in landslide susceptibility map generation by using the GBDT, RF and InV models. All of the maps of the causative factors were resampled to a resolution of 28.5 m. Of the 486289 pixels in the area,28526 pixels were landslide pixels, and 457763 pixels were non-landslide pixels. Finally, landslide susceptibility maps were generated by using the three machine learning models, and their performances were assessed through receiver operating characteristic(ROC) curves, the sensitivity, specificity,overall accuracy(OA), and kappa coefficient(KAPPA). The results showed that the GBDT, RF and In V models in overall produced reasonable accurate landslide susceptibility maps. Among these three methods, the GBDT method outperforms the other two machine learning methods, which can provide strong technical support for producing landslide susceptibility maps in TGR.

Three-dimensional conjugate gradient inversion of magnetotelluric full information data

基于阻抗张肌数据， tipper 数据，和conjugate 坡度算法的分析，我们发展一三维( 3D )为转换结合坡度算法完整的信息数据决定了从的 magnetotelluric 五电并且磁场部件并且讨论方法为 3D 倒置结果的量的解释使用完整的信息数据。从合成数据的 3D 倒置的结果显示从转换的结果更好联合张肌和 tipper 数据是的阻抗比的完整的信息数据源于转换仅仅阻抗张肌数据(或 tipper 数据) 在改进分辨率和可靠性。合成例子也表明这个 3D 倒置算法的有效性和稳定性。

Convergence analysis on Browder-Tikhonov regularization for second-order evolution hemivariational inequality

This paper studies the Browder-Tikhonov regularization of a second-order evolution hemivariational inequality （SOEHVI） with non-coercive operators. With duality mapping, the regularized formulations and a derived first-order evolution hemivariational inequality （FOEHVI） for the problem considered are presented. By applying the Browder-Tikhonov regularization method to the derived FOEHVI, a sequence of regularized solutions to the regularized SOEHVI is constructed, and the strong convergence of the whole sequence of regularized solutions to a solution to the problem is proved.

SINE TRANSFORM PRECONDITIONERS FOR SECOND-ORDER PARTIAL DIFFERENTIAL EQUATIONS

In this paper, we are concerned with the numerical solution of second-order partial differential equations. We analyse the use of the Sine Transform precondilioners for the solution of linear systems arising from the discretization of p.d.e. via the preconditioned conjugate gradient method. For the second-order partial differential equations with Dirichlel boundary conditions, we prove that the condition number of the preconditioned system is O(1) while the condition number of the original system is O(m 2) Here m is the number of interior gridpoints in each direction. Such condition number produces a linear convergence rale.

基于改进Informer的蒸汽管网流量预测

提出了一种基于改进Informer的蒸汽管网流量预测模型Informer_BEST。该模型通过引入扩展的因果卷积、相对位置和季节特征编码以及梯度中心化技术等3种算法,对Informer模型进行了改进和优化。对5种预测模型进行实验,结果表明,所提出的Informer_BEST预测模型展现了令人满意的测试结果,预测曲线与实际数据的变化趋势相吻合,对供热调度决策具有良好的指导作用。

基于算力-能量全分布式在线共享的5G网络负荷管理策略

5G与边缘计算等信息基础设施海量部署造成运营商用电成本上升,需推动边缘网络与电网的能量互动以节能降本。现有研究重点关注边缘网络参与日前经济调度,未考虑可再生能源和网络流量双重随机性造成的网络能量供需不平衡问题。针对强随机环境下的网络负荷管理问题,提出面向虚拟化边缘网络的能量实时管理策略。首先,以网络用能成本最小化为目标,构建联合网络资源管理、储能充放电与能量共享模型。其次,针对未来网络信息未知无法直接求解的问题,提出基于随机对偶次梯度法的在线管理策略。然后,针对资源共享涉及运营商隐私问题,提出全分布式的计算资源与能量协同共享算法。最后,仿真验证表明,所提在线算法在无需先验知识的前提下有效减少了5G边缘网络的用能成本。

多导联心电图识别的稳定步长ResNet深度网络

针对经典的ResNet深度神经网络对一维多导联心电图图像进行识别、分类时,因原始图像的维度较高导致提取到的深度特征维度高,造成全连接层训练出现收敛速度慢和过拟合的问题,在ResNet的全连接层提出一种稳定步长动量训练算法,通过引入近似二阶梯度信息增强动量法的寻优能力和加速收敛速度;利用连续2次迭代的参数变化量和梯度信息自适应调整步长,构造边界函数对步长的大小进行限制,以防止步长过大或过小而影响收敛稳定性,使用动量项对参数的更新方向进行修正。在CPSC2018心电图数据集上的实验结果表明:所提算法训练的ResNet取得的F 1分数、准确率、精确度分别达到0.859、97.4%、87.9%,收敛速度和整体分类指标值优于其他相比较的方法。

Data-Driven Learning Control Algorithms for Unachievable Tracking Problems

For unachievable tracking problems, where the system output cannot precisely track a given reference, achieving the best possible approximation for the reference trajectory becomes the objective. This study aims to investigate solutions using the Ptype learning control scheme. Initially, we demonstrate the necessity of gradient information for achieving the best approximation.Subsequently, we propose an input-output-driven learning gain design to handle the imprecise gradients of a class of uncertain systems. However, it is discovered that the desired performance may not be attainable when faced with incomplete information.To address this issue, an extended iterative learning control scheme is introduced. In this scheme, the tracking errors are modified through output data sampling, which incorporates lowmemory footprints and offers flexibility in learning gain design.The input sequence is shown to converge towards the desired input, resulting in an output that is closest to the given reference in the least square sense. Numerical simulations are provided to validate the theoretical findings.

水下重力匹配导航适配性评价方法比较研究

重力匹配导航精度受重力适配区选取的影响。针对如何有效筛选最优适配区,采用信息熵的单特征准则与多属性决策的多特征准则评价重力底图的适配性。选择重力标准差、相关系数、粗糙度、坡度、重力特征丰度与累加梯度均值6种参数,利用单/多特征两种准则筛选同一海域适配/非适配区,通过地形轮廓匹配(terrain contour matching, TERCOM)算法在所筛选区域进行重力匹配导航验证,对比分析两者的匹配导航精度。结果表明,单特征准则在适配/非适配区的导航误匹配率分别为30%和20%;多特征准则在适配区无明显误匹配,误差在百米级,在非适配区出现多次误匹配。分析表明,多特征分析准则与单特征分析准则相比,可更准确有效地评价重力匹配区的适配性。

管理层讨论与分析能预示企业违约吗?——基于中国股市的实证分析

采用文本挖掘技术,对上市公司年报中的管理层讨论与分析(MD&A)内容进行文本分析,从文本相似度、文本可读性、文本语调以及管理层预期的角度构建了MD&A评价体系。通过构建代价敏感GBDT(csGBDT)模型,考察多维管理层讨论与分析指标对企业违约预测的影响,并进一步分析了对企业违约状态有重要影响的MD&A指标及其对违约状态作用的边际效应。研究表明:MD&A指标可以作为替代性数据源准确预测上市公司违约状态;MD&A指标相比传统违约预测变量的预测效果较差;MD&A指标在传统违约判别指标基础上提供了额外的信息含量;csGBDT模型显著提高了对企业(尤其是对违约企业)的判别能力,在违约预测的大数据方法中具有明显优势。在众多管理层讨论与分析指标中,对企业违约有重要影响的MD&A指标依次为:与前一年相比文本相似度、词汇总量、情感语调2、词汇总量/句子数量、情感语调1和管理层是否发出业绩预测。本文将企业违约预测的研究边界从结构化数据拓展到非结构化文本数据,有助于抑制信息不对称导致的企业违约风险。

改进数字滤波的网络终端信息防泄漏方法仿真

网络终端信息中含有大量噪声,加大了网络信息泄露风险,为了有效防止网络终端信息泄露,提高信息安全性,提出改进数字滤波的网络终端信息防泄漏方法。以网络终端信息作为原点,在设定的象限内将信号点到当前点的欧式距离作为权值,对信号点作加权平均处理,获取网络终端信息中心位置坐标值。通过城区距离对窗口进行扩展,将拓展后的窗口作为滤波窗口对各个信号点加权处理,完成网络终端信息滤波处理,降低噪声对信息安全的影响。采用分布式梯度密钥管理策略获取最优值,将其设定为网络节点的私有密钥,组建密钥对。通过密钥对对采集的全部网络终端信息签名和加密处理,最终实现网络终端信息防泄漏处理。实验结果表明,所提方法可以有效确保网络终端信息安全性,同时还能够有效减少空间占用以及能耗。

基于关联信息熵和轻量级梯度提升机的油纸绝缘特征优选策略

为了充分挖掘对变压器油纸绝缘综合诊断结果有利的老化特征量,提出一种基于关联信息熵和轻量级梯度提升机(LightGBM)的特征量优选策略。首先,基于不同老化状态的变压器介电响应实测数据,提取不同类别的时域特征量形成初始高维特征空间;其次,引入关联信息熵度量特征子集的相关性及冗余性,再利用轻量级梯度提升机评估特征的重要度,进而得到最优特征空间;最后,对比分析最优特征空间与不同对照组的诊断性能,有效验证了基于所提优选策略确定的最优特征空间的优越性。

区域快速优化下的无人机在线轨迹规划方法

针对传统基于梯度的规划方法需预先构建欧式符号距离场(euclidean signed distance field, ESDF)导致障碍物信息冗余度高、规划效率受限问题,提出了一种基于区域快速优化的实时轨迹规划方法。所提方法设计碰撞控制点替换策略用于加快碰撞区域的轨迹优化收敛速度从而降低轨迹规划时间,并定义提取与轨迹规划相关的局部障碍物信息方法,避免构建ESDF过程,从而提高规划效率;之后考虑轨迹安全性、平滑性及动态可行性,建立多目标优化函数,进一步优化轨迹。仿真实验表明,该方法可有效实现无人机在线轨迹规划,且与前沿方法相比,轨迹规划时间平均缩短了36.1%,轨迹优化收敛速度平均提高了33.1%,实现了更高效的规划。

基于PSO-BP的抗乳腺癌药物毒性研究

为解决新药研发过程中药物的毒性难以准确预估的问题,利用计算机技术,提出一种基于粒子群算法(PSO)优化BP神经网络的二分类预测模型。通过互信息的方法从729个分子描述符中筛选出重要度最高的20特征作为自变量,以药物的毒性值作为因变量,在BP神经网络模型的基础上,首先使用不同的梯度下降算法计算模型的准确率,发现批量梯度下降算法对BP模型的拟合效果较好;其次利用动态变权重的粒子群算法对BP神经网络模型的权重和阈值进行优化选择,结合BP神经网络、SVM和KNN模型进行对比实验,结果显示,PSO-BP模型的准确率、精确率、召回率和F1值明显高于其它模型。因此,PSO-BP模型是一种对抗乳腺癌药物毒性有效预测的方法。

基于注意力机制的信息预处理多智能体强化学习算法

多智能体强化学习在群体控制领域具有广泛应用,然而传统的强化学习方法(如Q-Learning或策略梯度)在多智能体环境中表现不佳。在训练过程中,每个智能体的策略不断变化。当一个智能体基于环境信息做出决策时,其他智能体的决策可能已经影响了环境信息,导致智能体感知的转移概率分布和奖赏函数发生变化,使得环境变得非平稳,训练无法有效进行。为了缓解这一问题,研究了一种基于多头自注意力的多智能体强化学习算法。该方法考虑了其他智能体的行动策略,利用多头自注意力算法使智能体能够学习对决策影响最大的因素,成功地学习了复杂的多智能体协调策略。在实验结果中平均回报达值到了0.82,远高于传统算法的表现。实验结果表明,所提出的基于多头自注意力的多智能体强化学习算法能够有效解决环境不平稳导致的多智能体学习困难问题,提高了多智能体强化学习算法的收敛速度和平稳性。

基于MI-CEEMDAN-RF-LGBM的风功率预测分析

针对传统的风功率预测精度无法满足现场实际需求的问题,提出了一种基于互信息(MI)-完全自适应噪声集合经验模态分解(CEEMDAN)-随机森林(RF)-轻量级梯度提升机(LGBM)多种算法融合的风功率预测模型。采用MI法对风向、风速、温度等一系列风机参数数据进行特征选择,选出与风功率强相关的参数变量;利用CEEMDAN算法对原始风功率序列进行特征分解,将其分解成多个模态分量;为了防止建模输入过多造成数据的冗余,采用RF算法进行二次特征选择,对提取出的特征变量进行特征选择,进一步筛选出与风功率原始序列具有较高相关性的特征变量;利用LGBM算法、极限学习机(ELM)以及深度信念网络(DBN)分别建立风向预测模型,选择出建模精度更高的风功率预测模型。采用桂林某风电场53 747组、每组间隔为10min的风功率、风向、风速等风机参数数据进行试验,验证了所载模型的有效性。

Adiabatic Shear Localization for Steels Based on Johnson-Cook Model and Second-and Fourth-Order Gradient Plasticity Models

To consider the effects of the interactions and interplay among microstructures, gradient-dependent models of second- and fourth-order are included in the widely used phenomenological Johnson-Cook model where the effects of strain-hardening, strain rate sensitivity, and thermal-softening are successfully described. The various parameters for 1006 steel, 4340 steel and S-7 tool steel are assigned. The distributions and evolutions of the local plastic shear strain and deformation in adiabatic shear band （ASB） are predicted. The calculated results of the second- and fourth- order gradient plasticity models are compared. S-7 tool steel possesses the steepest profile of local plastic shear strain in ASB, whereas 1006 steel has the least profile. The peak local plastic shear strain in ASB for S-7 tool steel is slightly higher than that for 4340 steel and is higher than that for 1006 steel. The extent of the nonlinear distribution of the local plastic shear deformation in ASB is more apparent for the S-7 tool steel, whereas it is the least apparent for 1006 steel. In fourth-order gradient plasticity model, the profile of the local plastic shear strain in the middle of ASB has a pronounced plateau whose width decreases with increasing average plastic shear strain, leading to a shrink of the portion of linear distribution of the profile of the local plastic shear deformation. When compared with the sec- ond-order gradient plasticity model, the fourth-order gradient plasticity model shows a lower peak local plastic shear strain in ASB and a higher magnitude of plastic shear deformation at the top or base of ASB, which is due to wider ASB. The present numerical results of the second- and fourth-order gradient plasticity models are consistent with the previous numerical and experimental results at least qualitatively.

Underwater Terrain Image Stitching Based on Spatial Gradient Feature Block

At present,underwater terrain images are all strip-shaped small fragment images preprocessed by the side-scan sonar imaging system.However,the processed underwater terrain images have inconspicuous and few feature points.In order to better realize the stitching of underwater terrain images and solve the problems of slow traditional image stitching speed,we proposed an improved algorithm for underwater terrain image stitching based on spatial gradient feature block.First,the spatial gradient fuzzy C-Means algorithm is used to divide the underwater terrain image into feature blocks with the fusion of spatial gradient information.The accelerated-KAZE(AKAZE)algorithm is used to combine the feature block information to match the reference image and the target image.Then,the random sample consensus(RANSAC)is applied to optimize the matching results.Finally,image fusion is performed with the global homography and the optimal seam-line method to improve the accuracy of image overlay fusion.The experimental results show that the proposed method in this paper effectively divides images into feature blocks by combining spatial information and gradient information,which not only solves the problem of stitching failure of underwater terrain images due to unobvious features,and further reduces the sensitivity to noise,but also effectively reduces the iterative calculation in the feature point matching process of the traditional method,and improves the stitching speed.Ghosting and shape warping are significantly eliminated by re-optimizing the overlap of the image.

基于数据驱动的电力系统虚假数据注入攻击检测

针对虚假数据注入攻击(FDIA)导致的电力系统安全问题,提出了一种数据驱动的二阶段联合方法。首先,构建由贝叶斯优化(BO)改进的极端梯度提升算法BO-XGBoost作为基分类器对正常数据和虚假数据进行初筛。其次,考虑到分类问题有一定误报率,构建最大互信息系数(MIC)校验器,计算由基分类器判断为正常状态下系统节点间的MIC,基于该系数的值进一步校验潜藏在系统里的FDIA。最后,通过IEEE 39节点测试系统的仿真分析,验证了所提方法的有效性。

基于梯度分块的直方图平移可逆信息隐藏

针对大多数基于直方图平移的可逆信息隐藏中载体图像分块方法过于粗糙的问题,根据相邻像素之间具有相似特性,提出对载体图像按像素梯度进行分割,将图像分为平滑区域和粗糙区域。同时为减少像素的无效平移,只对平滑区域分块,并采用直方图的次峰值点进行双向移位实现信息可逆隐藏,同时将零值点嵌入图像;不需要单独传送峰值点和零值点信息,降低了通信负担。实验结果表明:所提方法在图像质量和峰值信噪比上具有明显的提升,可应用于保密传输等领域。