基于ST-ConvLSTM的南海海表面CO2分压的空间和时间序列预测

海表面二氧化碳分压(pCO₂)的未来变化趋势,对统计评估全球碳收支以及理解全球气候变化背景下的海洋酸化现象至关重要。目前传统的海面pCO₂预测方法大部分基于有限的实测数据,然而实测数据存在着时间和地理方面的制约,且计算成本较高。近年来,随着时空观测数据的爆炸性增长,基于深度学习的数据驱动模型在海表面pCO₂预测方面中表现出良好的潜力。然而,由于多种环境因素与海表面pCO₂之间的关系错综复杂,到目前为止尚无十分简单有效的相关模型来对海表面pCO₂进行预测。为应对这一挑战,利用时空卷积长短时记忆神经网络(ST-ConvLSTM)模型,通过海面温度(sea surface temperature, SST)、海面盐度(sea surface salinity, SSS)、叶绿素a浓度(chl a)和海面pCO₂数据,预测南海的海面pCO₂,并将2019年1～12月的数据作为测试集对模型的表现进行了验证。结果显示, ST-ConvLSTM模型...     

A novel hybrid approach of Bayesian Logistic Regression and its ensembles for landslide susceptibility assessment

Abstract

A novel artificial intelligence approach of Bayesian Logistic Regression (BLR) and its ensembles [Random Subspace (RS), Adaboost (AB), Multiboost (MB) and Bagging] was introduced for landslide susceptibility mapping in a part of Kamyaran city in Kurdistan Province, Iran. A spatial database was generated which includes a total of 60 landslide locations and a set of conditioning factors tested by the Information Gain Ratio technique. Performance of these models was evaluated using the area under the ROC curve (AUROC) and statistical index-based methods. Results showed that the hybrid ensemble models could significantly improve the performance of the base classifier of BLR (AUROC?=?0.930). However, RS model (AUROC?=?0.975) had the highest performance in comparison to other landslide ensemble models, followed by Bagging (AUROC?=?0.972), MB (AUROC?=?0.970) and AB (AUROC?=?0.957) models, respectively.     

深度学习PaddlePaddle框架支持下的遥感智能视觉平台研究与实现

百度深度学习PaddlePaddle框架支持下的遥感智能视觉平台,能够运用深度学习技术实现遥感影像的智能建模、训练和解译。本文通过深入分析PaddlePaddle图像分割模型库PaddleSeg的图像处理深度学习算法模型DeepLabV3+、U²-Net及RetinaNet,开发设计了遥感智能视觉平台,实现了遥感影像的地块分割、变化检测和斜框检测等专业功能。研究表明：遥感智能视觉平台提取的图斑总面积是目视解译的80%、有效图斑比例为76%、错误图斑比例为18%,实现了快速有效的遥感图像智能处理。     

Prediction of Stellar Atmospheric Parameters using Instance-Based Machine Learning and Genetic Algorithms

In this article we present a method for the automated prediction of stellar atmospheric parameters from spectral indices. This method uses a genetic algorithm (GA) for the selection of relevant spectral indices and prototypical stars and predicts their properties, using the k-nearest neighbors method (KNN). We have applied the method to predict the effective temperature, surface gravity, metallicity, luminosity class and spectral class of stars from spectral indices. Our experimental results show that the feature selection performed by the genetic algorithm reduces the running time of KNN up to 92%, and the predictive accuracy error up to 35%. This revised version was published online in July 2006 with corrections to the Cover Date.     

Integrating farmers’ adaptive knowledge into flood management and adaptation policies in the Vietnamese Mekong Delta: A social learning perspective

Flood management and adaptation are important elements in sustaining farming production in the Vietnamese Mekong Delta (VMD). While over the past decades hydraulic development introduced by the central government has substantially benefited the rural economy, it has simultaneously caused multiple barriers to rural adaptation. We investigate the relational practices (i.e., learning interactions) taking place within and across the flood management and adaptation boundaries from the perspective of social learning. We explore whether and how adaptive knowledge (i.e., experimental and experiential knowledge) derived from farmers’ everyday adaptation practices contributes to local flood management and adaptation policies in the selected areas. We collected data through nine focus groups with farmers and thirty-three interviews with government officials, environmental scientists, and farmers. Qualitative analysis suggests that such processes are largely shaped by the institutional context where the boundary is embedded. This study found that while the highly bureaucratic operation of flood management creates constraints for feedback, the more informal arrangements set in place at the local level provide flexible platforms conducive to open communication, collaborative learning, and exchange of knowledge among the different actors. This study highlights the pivotal role of shadow systems that provide space for establishing and maintaining informal interactions and relationships between social actors (e.g., interactions between farmers and extension officials) in stimulating and influencing, from the bottom-up, the emergence of adaptive knowledge about flood management and adaptation in a local context.     

多尺度字典学习的高光谱图像压缩算法

结合小波变换及字典学习提出了一种针对高光谱图像的压缩算法。该算法首先通过小波变换构建多尺度样本集,在小波域使用K-均值奇异值分解(K-SVD)方法学习得到原子尺寸不同的多尺度字典,然后在稀疏表示的过程中,定义一个原子使用频次筛选因子,通过统计局部最优波段稀疏表示时原子使用情况,结合筛选因子对字典原子进行优化筛选,使用精简后的字典对其余波段进行稀疏求解,最后针对不同尺度的表示系数采用自适应的量化编码。实验结果表明,与目前常用的3D-SPIHT和其他的多尺度字典学习算法相比,本文算法在中低比特率下,具有更好的重建性能。     

融合多模态特征的社会多媒体谣言检测技术研究

以微博为代表的社会媒体的蓬勃发展在加速信息交流的同时,也促使虚假谣言信息迅速在社会网络上传播,造成严重的后果.自动谣言检测问题受到了国内外学术界、产业界的广泛关注.围绕社会多媒体谣言检测这一问题,本文总结了融合多模态特征的谣言检测相关技术.首先从基本概念出发,阐述了谣言的定义和社会多媒体的特点,给出了社会多媒体谣言检测问题的定义.针对谣言检测面临的多模态特征抽取和模型构建两大难点,分别总结和归纳了各种类型的特征及其提取方法和不同的机器学习检测模型.这些特征和算法是检测谣言的基本手段,也是接下来研究的基础,可为进一步谣言检测的研究提供参考.     

Site characterization model using least‐square support vector machine and relevance vector machine based on corrected SPT data (Nc)

Statistical learning algorithms provide a viable framework for geotechnical engineering modeling. This paper describes two statistical learning algorithms applied for site characterization modeling based on standard penetration test (SPT) data. More than 2700 field SPT values (N) have been collected from 766 boreholes spread over an area of 220 sqkm area in Bangalore. To get N corrected value (N_c), N values have been corrected (N_c) for different parameters such as overburden stress, size of borehole, type of sampler, length of connecting rod, etc. In three‐dimensional site characterization model, the function N_c=N_c (X, Y, Z), where X, Y and Z are the coordinates of a point corresponding to N_c value, is to be approximated in which N_c value at any half‐space point in Bangalore can be determined. The first algorithm uses least‐square support vector machine (LSSVM), which is related to a ridge regression type of support vector machine. The second algorithm uses relevance vector machine (RVM), which combines the strengths of kernel‐based methods and Bayesian theory to establish the relationships between a set of input vectors and a desired output. The paper also presents the comparative study between the developed LSSVM and RVM model for site characterization. Copyright © 2009 John Wiley & Sons, Ltd.     

迁移学习支持下的高分影像山地滑坡灾害解译模型

无人机低空遥感是近年来新兴的一种快速获取灾情信息的手段,如何利用无人机高分影像构建滑坡灾害解译模型是实现快速自动解译滑坡的关键。针对该问题,对比了多种影像特征提取方法,将迁移学习(TL)特征和支持向量机(SVM)引入到构建滑坡灾害自动解译模型中,提出了一种TL支持下的高分影像滑坡灾害解译模型。选取5·12汶川地震及4·20芦山地震系列无人机影像构建了滑坡灾害样本库并进行了实验,TL特征方法整体分类准确度ACC为95%,ROC达到0.98,识别准确率达到97%。结果表明,所提方法可用于高分影像滑坡自动解译,同时可用于大面积高分影像中快速山地滑坡灾害定位及检测。     

历史数据和强化学习相结合的低频轨迹数据匹配算法

针对低频(采样间隔大于1min)轨迹数据匹配算法精度不高的问题,提出了一种基于强化学习和历史轨迹的匹配算法HMDP-Q,首先通过增量匹配算法提取历史路径作为历史参考经验库;根据历史参考经验库、最短路径和可达性筛选候选路径集;再将地图匹配过程建模成马尔科夫决策过程,利用轨迹点偏离道路距离和历史轨迹构建回报函数;然后借助强化学习算法求解马尔科夫决策过程的最大回报值,即轨迹与道路的最优匹配结果;最后应用某市浮动车轨迹数据进行试验。结果表明:本文算法能有效提高轨迹数据与道路匹配精度;本算法在1min低频采样间隔下轨迹匹配准确率达到了89.2%;采样频率为16min时,该算法匹配精度也能达到61.4%;与IVVM算法相比,HMDP-Q算法匹配精度和求解效率均优于IVVM算法,16min采样频率时本文算法轨迹匹配精度提高了26%。