共查询到20条相似文献，搜索用时 62 毫秒
1.
Traffic transformer: Capturing the continuity and periodicity of time series for traffic forecasting
Traffic forecasting is a challenging problem due to the complexity of jointly modeling spatio‐temporal dependencies at different scales. Recently, several hybrid deep learning models have been developed to capture such dependencies. These approaches typically utilize convolutional neural networks or graph neural networks (GNNs) to model spatial dependency and leverage recurrent neural networks (RNNs) to learn temporal dependency. However, RNNs are only able to capture sequential information in the time series, while being incapable of modeling their periodicity (e.g., weekly patterns). Moreover, RNNs are difficult to parallelize, making training and prediction less efficient. In this work we propose a novel deep learning architecture called Traffic Transformer to capture the continuity and periodicity of time series and to model spatial dependency. Our work takes inspiration from Google’s Transformer framework for machine translation. We conduct extensive experiments on two real‐world traffic data sets, and the results demonstrate that our model outperforms baseline models by a substantial margin. 相似文献
2.
Abstract Conventional methods of deriving global or continental vegetation maps from the National Oceanic and Atmospheric Administration's (NOAA) Advanced Very High Resolution Radiometer (AVHRR) time series data are based on two‐value Boolean logic, which cannot properly model the so‐called ecotone, the transition zone between adjacent ecosystems. New methods and data models that have been developed on the basis of fuzzy logic to address the “mixed pixel” issue in multi‐spectral imagery can also be used with multi‐temporal imagery to handle the mixture of vegetation types within an ecotone. This study introduces the concept of semantic space and its transformation from spectral feature space, which utilizes a fuzzy logic approach to characterize the continuum of vegetation communities in the African continent from AVHRR multi‐temporal (12 months for three years from 1986 to 1988) NDVI data. The fuzzy procedure was based on the Fuzzy c‐Means (FCM) algorithm with significant modifications to improve processing speed for handling large volumes of data. A second‐order mapping approach was also devised to explicitly represent subdominant vegetative coverage in ecotones and other heterogeneous regions. Comparisons between a Sub‐Saharan African Vegetation Map compiled by the International Union for Conservation of Nature (IUCN) in 1986 and the maps derived from this study demonstrated that fuzzy modeling and classification might provide a better and more realistic representation of the vegetative characteristics of the region. 相似文献
3.
随着对GIS中的空间对象模型和自然地理特征表达的研究深入,模糊空间对象被提出。针对模糊空间对象表达的特点,提出了一种基于模糊神经网络的模糊空间对象生成方法。该方法将模糊技术与神经网络相结合,利用神经网络的学习能力调整模糊隶属函数和模糊规则,使系统具备自适应的特性。实验表明,这种基于模糊神经网络的生成模糊空间对象的方法比传统方法大大的提高了成果的精度。 相似文献
4.
基于RNN的空气污染时空预报模型研究 总被引：2，自引：0，他引：2
针对空气污染物时间序列中包含缺失值以及现有时间序列预报模型缺乏对时序特征状态建模的问题,该文构建了基于缺失值处理算法和RNN(循环神经网络)的时空预报框架。对空气污染物时序数据设计了3种缺失值处理算法(前向递补、均值替代和权重衰减),用缺失标签和缺失时长对缺失值建模,并在此基础上搭建含有全连接层与LSTM层的深度循环神经网络(DRNN)用于时空预报。使用深度全连接神经网络(DFNN)作为DRNN的对照,用京津冀区域的空气质量和气象数据训练模型,并比较不同模型的预测精度。通过实验,比较了3种缺失值处理方法的效果,结果表明,LSTM在空气污染时空序列预测上的表现优于传统的全连接神经网络层,证实了提出的基于深度学习的时空预报框架的有效性。 相似文献
5.
随着对GIS中的空间对象模型和自然地理特征表达的研究深入,模糊空间对象被提出。针对模糊空间对象表达的特点,提出了一种基于模糊神经网络的模糊空间对象生成方法。该方法将模糊技术与神经网络相结合,利用神经网络的学习能力调整模糊隶属函数和模糊规则,使系统具备自适应的特性。实验表明,这种基于模糊神经网络的生成模糊空间对象的方法比传统方法大大的提高了成果的精度。 相似文献
6.
Simon Corne Tavi Murray Stan Openshaw Linda See Ian Turton 《Journal of Geographical Systems》1999,1(1):37-60
Measurements of water pressure beneath Trapridge Glacier, Yukon Territory, Canada show that the basal water system is highly
heterogeneous. Three types of behaviour were recorded: pressure records which are strongly correlated, records which are strongly
anticorrelated, and records which alternate between strong correlation and strong anticorrelation. We take the pressure in
bore-holes that are connected to the evacuation route for basal water as the forcing, and the other pressures as the response
to this forcing. Previous work (Murray and Clarke 1995) has shown that these relationships can be modelled using low-order
nonlinear differential equations optimized by inversion. However, despite optimizing the model parameters we cannot be sure that the final model forms are themselves optimal. Computational intelligence techniques provide alternative methods for fitting models and are robust
to missing or noisy data, applicable to non-smooth models, and attempt to derive optimal model forms as well as optimal model
parameters. Four computational intelligence techniques have been used and the results compared with the more conventional
mathematical model. These methods were genetic programming, artificial neural networks, fuzzy logic and self-organizing maps.
We compare each technique and offer an evaluation of their suitability for modelling the pressure data. The evaluation criteria
are threefold: (1) goodness of fit and an ability to predict subsequent data under different surface weather conditions; (2)
interpretability, and the extent and significance of any new insights offered into the physics of the glacier; (3) computation
time. The results suggest that the suitability of the computational intelligence techniques to model these data increases
with the complexity of the system to be modelled. 相似文献
7.
基于计算智能的土地适宜性评价模型 总被引：22，自引：2，他引：22
将计算智能理论引入土地评价领域，构建了一个全新的土地适宜性评价模型。首先基于模糊逻辑和人工神经网络构造了一个模糊神经网络模型，然后采用改进的遗传算法进行训练，能够快速收敛到最优解，对初始的规则库进行修正，形成了一个自学习、自适应的评价系统。 相似文献
8.
针对无人机影像深度学习分类方法缺乏现状,本文利用深度学习理论卷积神经网络方法对无人机影像进行了分类。该法首先抽取无人机影像作为训练集和检验集,然后建立一个2个卷积层-池化层的卷积神经网络模型进行深度学习,通过设定参数并运行模型实现无人机影像分类。实验表明,本文提出的方法可完成较复杂地区无人机影像分类,其分类精度与支持向量机方法相当,为无人机遥感影像分类提供了一个崭新的技术视点。 相似文献
9.
JIAO Limin LIU Yaolin 《地球空间信息科学学报》2007,10(2):151-156
A novel model of land suitability evaluation is built based on computational intelligence (CI). A fuzzy neural network (FNN) is constructed by the integration of fuzzy logic and artificial neural network (ANN). The structure and process of this network is clear. Fuzzy rules (knowledge) are expressed in the model explicitly, and can be self-adjusted by learning from samples. Genetic algorithm (GA) is employed as the learning algorithm to train the network, and makes the training of the model efficient. This model is a self-learning and self-adaptive system with a rule set revised by training. 相似文献
10.
基于神经网络的地图数字注记识别 总被引：1，自引：0，他引：1
指出了地图自动识别系统中地图数字注记识别存在的困难，论证了利用神经网络技术解决这种困难的可能性，并通过一个含有２个隐藏层的ＢＰ网络，说明了这种技术用于地图数字注记识别的可行性。 相似文献
11.
本文概述了神经网络计算机的基本特点、神经网络的形式化描述、目前国际上神经网络计算机的现况以及在遥感图像处理中应用的潜力和展望。 相似文献
12.
深度学习技术促使诸多领域研究取得突破性进展, 基于深度神经网络的地图综合研究备受期待。将建筑物综合过程抽象解释为编解码过程, 系统地研究基于编解码结构的深度神经网络在建筑物综合中的应用。首先, 利用空间划分与矢量-栅格数据转换相结合的方式构建样本和样本集; 然后, 利用样本集训练基于编解码结构的深度神经网络, 实现建筑物综合学习泛化并测试、评估其效果; 最后, 搭建5种代表性的基于编解码结构的深度神经网络, 分析比较各模型在建筑物综合中的应用效果。实验结果表明, 基于编解码结构的深度神经网络能够从建筑物综合样本中学习或推理出部分建筑物综合知识和综合操作, 且5种模型中Pix2Pix更适用于建筑物综合的学习模拟。 相似文献
13.
Manfred M Fischer 《Transactions in GIS》2006,10(4):521-533
The focus of this paper is on the neural network modelling approach that has gained increasing recognition in GIScience in recent years. The novelty about neural networks lies in their ability to model non‐linear processes with few, if any, a priori assumptions about the nature of the data‐generating process. The paper discusses some important issues that are central for successful application development. The scope is limited to feedforward neural networks, the leading example of neural networks. It is argued that failures in applications can usually be attributed to inadequate learning and/or inadequate complexity of the network model. Parameter estimation and a suitably chosen number of hidden units are, thus, of crucial importance for the success of real world neural network applications. The paper views network learning as an optimization problem, reviews two alternative approaches to network learning, and provides insights into current best practice to optimize complexity so to perform well on generalization tasks. 相似文献
14.
当某一问题很难甚至无法用数学方法建立精确模型时,人工神经网络的方法则显示了优势。对于一个具体问题,采用何种网络结构是至关重要的。本文以美国内华达州Cuprite矿区成像光谱数据特征矿物识别为例,采用6种不同结构的多层前馈网络模型,从其训练难度、运算效率及识别效果等方面进行了综合对比分析。 相似文献
15.
本文在研究BP神经网络和模糊理论的基础上,提出了传统BP算法的一种改进方法和基于模糊系统的神经网络遥感影像分类方法。通过试验表明:基于模糊技术的神经网络分类方法要优于BP神经网络方法,取得了令人满意的效果。 相似文献
16.
Elahe Khesali Mohammad Javad Valadan Zoej Mehdi Mokhtarzade Maryam Dehghani 《Journal of the Indian Society of Remote Sensing》2016,44(1):21-29
Two new methods for fusion of high-resolution optical and radar satellite images have been proposed to extract roads in high quality in this paper. Two fusion methods, including neural network and knowledge-based fusion are introduced. The first proposed method consists of two stages: (i) separate road detection using each dataset and (ii) fusion of the results obtained using a neural network. In this method, the neural networks are separately applied on high-resolution IKONOS and TerraSAR-X images for road detection, using a variety of texture parameters. The outputs of two neural networks, as well as the spectral features of optical image, are used in a third neural network as inputs. The second method is a knowledge-based fusion using thresholds of narrow roads and vegetation gray levels. First roads are extracted from each source separately. The outputs are then compared and advantages and disadvantages of each data source are investigated . The results obtained from accuracy assessment show the efficiency of the proposed methods. Furthermore, the comparison of the results showed the superiority of the first algorithm. 相似文献
17.
A. K. Mitra Sankar Nath A. K. Sharma 《Journal of the Indian Society of Remote Sensing》2008,36(3):243-253
Operational meteorology is perceived as a fuzzy environment in which information is vaguely defined. The mesoscale processes
such as fog, stratus and convection are generally dependent on the topography of the place and has always been difficult to
forecast for the meteorologists. The main objective of the present study is to introduce the concept of fuzzy inference system (FIS) in the prediction of fog. This approach uses the concept of a pure fuzzy logic system where the fuzzy rule base
consists of a collection of fuzzy IF-THEN rules. The fuzzy inference engine uses these fuzzy IF-THEN rules to determine a mapping from fuzzy sets in the input universe of discourse to fuzzy sets in the output universe of discourse
based on fuzzy logic principles. Basic weather elements, which affect weather characteristics of fog, are fuzzified. These
are then used in fuzzy weather prediction models based on fuzzy inferences. These models are simulated and the crisp results
obtained using developed defuzzification strategies are compared with the actual weather data. The basis of methodology is
to construct the fuzzy rule base domain from the available daily current weather observations in winter season over New Delhi.
The results reveal that dew point spread and rate of change of dew point spread are the most important parameters for the
formation of fog. The results further indicate that fog formation over New Delhi are dominant when (i) dew point is greater
then 7°C along with dew point spread between 1 and 3°C. (ii) rate of change of dew point spread must be negative and wind
speed should be less than 4 knots. This study presents a technique for predicting the probability of fog over New Delhi for
5–6 hours in advance. The skill score indicates that the performance of FIS is appreciably good. The method is found to be
promising for operational application. 相似文献
18.
19.
《International Journal of Digital Earth》2013,6(4):343-358
Abstract The Digital Earth concept has attracted much attention recently and this approach uses a variety of earth observation data from the global to the local scale. Imaging techniques have made much progress technically and the methods used for automatic extraction of geo-ralated information are of importance in Digital Earth science. One of these methods, artificial neural networks (ANN) techniques, have been effectively used in classification of remotely sensed images. Generally image classification with ANN has been producing higher or equal mapping accuracies than parametric methods. Comparative studies have, in fact, shown that there is no discernible difference in classification accuracies between neural and conventional statistical approaches. Only well designed and trained neural networks can present a better performance than the standard statistical approaches. There are, as yet, no widely recognised standard methods to implement an optimum network. From this point of view it might be beneficial to quantify ANN's reliability in classification problems. To measure the reliability of the neural network might be a way of developing to determine suitable network structures. To date, the problem of confidence estimation of ANN has not been studied in remote sensing studies. A statistical method for quantifying the reliability of a neural network that can be used in image classification is investigated in this paper. For this purpose the method is to be based on a binomial experimentation concept to establish confidence intervals. This novel method can also be used for the selection of an appropriate network structure for the classification of multispectral imagery. Although the main focus of the research is to estimate confidence in ANN, the approach might also be applicable and relevant to Digital Earth technologies. 相似文献
20.
合成孔径雷达SAR（Synthetic Aperture Radar）是开展城市建筑区信息获取与动态监测的重要数据源。本文建立了一个面向深度学习建筑区提取的中高分辨率SAR建筑区数据集SARBuD1.0 （SAR BUilding Dataset）。该数据集包含了覆盖中国不同区域的27景高分三号（GF-3）精细模式SAR图像,并从中获取了建筑区共计60000个SAR样本数据,结合光学图像与专家解译,制作了与样本数据对应的标签图像。SARBuD1.0数据集包含了不同地形场景类型、不同分布类型、不同区域的建筑区。该数据集可支持研究者对建筑区进行图像特征分析、辅助图像理解,并可对当前热点深度学习方法提供训练、测试数据支持。本文以山区建筑为例,使用传统纹理特征与深度学习特征对建筑区进行了特征分析与比较,相比于传统的人工设计的纹理特征,卷积神经网络具有更深、更多的特征,利用网络模型浅层的不同卷积核采样可得到各种纹理特征,在网络的深层卷积结构中可获取代表着类别的深层语义特征,使得分类器能更好地检测并提取图像中指定的目标。基于本数据集利用深度学习方法对不同地形区域的建筑区进行提取实验。实验结果表明基于本数据集训练的深度学习模型,对建筑区提取可以取得良好的结果,说明该数据集可以很好支持面向大数据的深度学习方法。其他学者可以基于SARBuD1.0数据集开展建筑区图像特征分析与语义分割提取等方面的研究。 相似文献