模块化模糊神经网络训练策略研究

为分析模块化模糊神经网络的训练机制,进一步降低其训练复杂度,增强模型的泛化性能,尝试采用部分训练样本对模型进行相对粗训练,并以广西全省1957-2003年连续47 a 5月平均降水量作为实验数据,统计各种参数条件下模型的逐步预测误差情况.结果表明:由部分训练样本参与训练的模型所得测试误差普遍低于由全部训练样本参与训练的模型所得测试误差,证明该方法可行.     

Effects of local site and building parameters on damage distribution in Angra do Heroísmo—Azores

During the M_w=6.9 January 1, 1980 earthquake in the Azores, the damage distribution in the town of Angra do Heroísmo, in Terceira Island, had a non-homogeneous spatial character. In order to understand the observed pattern of damage a microtremor survey was carried out, using ambient vibrations, and the data were analysed with the J-Sesame software, following the H/V Nakamura methodology. The results are in good agreement with the surface geology and they show a good correlation with the observed damages. Two building surveys were performed, in 1980 on damage evaluation and in 2000 on characterization of few building parameters. With all these elements (surface geology type, soil peak frequency, building type, number of stories and damage level) for 2111 buildings it was possible to perform several interparameter analysis, which enabled the identification of indicators that could be used to improve the damage estimation for future events.     

Attenuation in the uppermost inner core from PKP recordings at African seismological stations

The attenuation factor Q_P at the top of the inner core is evaluated by using the amplitude spectral ratio of PKPdf and PKPbc phases observed at African stations (BGCA mostly), from strong deep earthquakes in the Pacific Ocean area. The maximum depth of penetration of the PKPdf phase into the inner core (IC) is roughly 377 km, and the sampled region of IC is centered beneath the Southern Indian Ocean. The derived mean value of Q_P is 249 ± 31 (95% confidence level) in the frequency range 0.2–2 Hz, where no frequency dependence of attenuation has been reliably observed. By using Student’s t-test, we show that the value is statistically significantly different (with a probability greater than 95%) from other mean values of Q derived by using the same method, for both the western (180 °W to 40 °E) and eastern (40 °E to 180 °E) hemispheres of the IC. The decrease of Q with the radius of the turning point (denoted by r^TP), according to Q_P = 840 − 0.62 r^TP, has a moderate statistical support (the R-squared value is 38%). A slightly increase of Q as a function of the angle of the PKPdf path within the inner core with respect to the Earth’s spin axis is observed, in agreement with various investigations performed in the time domain. However, the value of the anisotropy, if any, is suggested to be around 3%.     

基于标识的遥感影像部分监督模糊聚类

根据遥感影像的特点引入了部分监督模糊聚类的方法,并对其进行改进,加入了标识参数flagk,使其在训练过程中对参与计算的数据进行筛选,使得训练的速度加快,去除了非样本类数据对类心数据的影响,获得了较高精度的分类结果。     

基于自动搜索和光谱匹配技术的训练样本纯化算法

提出了一种基于局部自动搜索和光谱匹配技术的监督分类训练样本的纯化方法。该方法首先利用遥感影像中像元的灰度信息在图像上局部范围内自动搜索和选择最佳样区位置，然后利用光谱匹配的思想对寻找到的最佳样区在光谱空间上进一步纯化。实验结果证明，通过手工选择样区的辅助，该算法能够自动有效地搜寻到最佳样区的位置，并对最佳样区进行纯化处理。原始遥感图像经过本文的样区纯化算法处理后，无论是目视判读效果，还是分类后混淆矩阵的统计及分类精度，均优于纯化处理前的分类结果，具有一定的实用价值。     

IGS测站的非线性变化研究

基于坐标模式的广义网平差模型，利用IGS发布的对GPS全球站处理后形成的单天解SINEX文件，通过自编软件计算了IGS测站的时间序列，并发现IGS测站存在非线性变化。利用频谱分析方法得出了IGS测站存在年周期或半年周期的变化，同时利用经验模型建立了IGS测站的测站速度函数。指出了仅利用1个线性量估计IGS跟踪站的速度值会存在偏差，应建立周期性变化模型，或采用分段线性化的方法，每隔一定的时间给出1个对应的速度。     

Reliability of Ambient Noise Horizontal-to-Vertical Spectral Ratio in Urban Environments: The Case of Thessaloniki City (Northern Greece)

In the framework of an ambient noise measurement project in the downtown district of the City of Thessaloniki (Northern Greece), several tests and evaluation of noise recordings were performed in selected sites. This data set was processed and compared with previous results for the same sites in terms of horizontal-to-vertical (H/V) spectral ratio obtained from weak and strong motion earthquake data. The recording system used was tested against a reference one. Finally, systematic ambient noise measurements were performed at eight sites in the citys downtown area, to evaluate diurnal and seasonal variations. It was concluded that the recording system used was reliable and the ambient noise H/V spectral ratios were in good agreement with relevant results previously obtained. Diurnal variation of the ambient noise H/V spectral ratio showed that it is preferable to perform measurements during the calm hours of the day, when man-made noise is relatively low. However, no systematic seasonal fluctuation effect on the ambient noise H/V spectral ratio was identified.     

基于最短距离的细胞图像分离

在细胞图像的分析处理中,经常会遇到细胞聚堆的问题,即多个细胞粘连、重叠在一起.为了进一步的分析研究,需要将聚堆细胞分离成单个细胞,虽然已有很多比较成熟的算法,但几乎每一种算法都有其局限性.为了解决这一问题,提出了一种基于面积和周长相结合,外加形状因子为约束条件的最短距离的分离算法来进行分离.其中主要叙述了算法设计思想及其实现,实验结果证明算法是有效的.     

Spatio-temporal pattern of viral meningitis in Michigan, 1993-2001

To characterize Michigans high viral meningitis incidence rates, 8,803 cases from 1993–2001 were analyzed for standard epidemiological indices, geographic distribution, and spatio-temporal clusters. Blacks and infants were found to be high-risk groups. Annual seasonality and interannual variability in epidemic magnitude were apparent. Cases were concentrated in southern Michigan, and cumulative incidence was correlated with population density at the county level (r=0.45, p<0.001). Kulldorffs Scan test identified the occurrence of spatio-temporal clusters in Lower Michigan during July–October 1998 and 2001 (p=0.01). More extensive data on cases, laboratory isolates, sociodemographics, and environmental exposures should improve detection and enhance the effectiveness of a Space-Time Information System aimed at prevention.The authors gratefully acknowledge Dr. Duane Newton, former manager of the Virology Section, Michigan Department of Community Health Bureau of Laboratories, for his provision of data on viral isolates. We also appreciate the contributions of Drs. Leah Estberg and Dunrie Greiling for ClusterSeer support, Scott Swan for his assistance with ArcView GIS, and Dr. Edward L. Ionides for his advice on time series analysis. We are grateful to Dr. Pierre Goovaerts and two anonymous reviewers for their insightful comments on an earlier draft of this paper. This project was supported in part by a grant from the National Oceanic and Atmospheric Administrations Joint Program on Climate Variability and Human Health, a consortium including the EPA, NASA, NSF, and EPRI (NA16GP2361) to MLW.     

Measuring similarity between geospatial lifelines in studies of environmental health

Many epidemiological studies involve analysis of clusters of diseases to infer locations of environmental hazards that could be responsible for the disease. This approach is however only suitable for sedentary populations or diseases with small latency periods. For migratory populations and diseases with long latency periods, people may change their residential location between time of exposure and onset of ill health. For such situations, clusters are diffused and diluted by in- and out-migration and may become very difficult to detect. One way to address the problem of diffused clusters is to include in analyses not only current residential locations, but all past locations at which cases might have been exposed to environmental hazardous. In this paper, we assume that a persons residential history provides such information and represent it through a discrete geospatial lifeline data model. Clusters of similar geospatial lifelines represent individuals who have similar residential histories—and therefore represent people who are more likely to have had similar environmental exposure histories. We therefore introduce a lifeline distance (dissimilarity) measure to detect clusters of cases, providing a basis for revealing possible regions in space-time where environmental hazards might have existed in the past. The ability of the measure to distinguish cases from controls is tested using two sets of synthetically generated cases and controls. Results indicate that the measure is able to consistently distinguish between populations of cases and controls with statistically significant results. The lifeline distance measure consistently outperforms another measure which uses only the distance between subjects residences at time of diagnosis. However, the advantages of using the entire residential history are only partly realized, since the ability to distinguish between cases and controls is only moderately better for the lifeline distance function. Future work is needed to investigate modifications to the inter-lifeline distance measure in order to enhance the potential of this approach to detect locations of environmental hazards over the lifespan.This project is supported by grant number 1 R01 ES09816-01 from the National Institute of Environmental Health Sciences, NIH. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the NIEHS or NIH. We wish to thank Peter Rogerson for helpful discussions of the migration models, and the anonymous reviewers for pointing out areas where the paper could be improved.