K-means聚类算法随机确定初始聚类数目,而且原始数据集中含有大量的冗余特征会导致聚类时精度降低,而布谷鸟搜索(CS)算法存在收敛速度慢和局部搜索能力弱等问题,为此提出一种基于自适应布谷鸟优化特征选择的K-means聚类算法(DCFSK)。首...K-means聚类算法随机确定初始聚类数目,而且原始数据集中含有大量的冗余特征会导致聚类时精度降低,而布谷鸟搜索(CS)算法存在收敛速度慢和局部搜索能力弱等问题,为此提出一种基于自适应布谷鸟优化特征选择的K-means聚类算法(DCFSK)。首先,为提升CS算法的搜索速度和精度,在莱维飞行阶段,设计了自适应步长因子;为调节CS算法全局搜索和局部搜索之间的平衡、加快CS算法的收敛,动态调整发现概率,进而提出改进的动态CS算法(IDCS),在IDCS的基础上构建了结合动态CS的特征选择算法(DCFS)。其次,为提升传统欧氏距离的计算精确度,设计同时考虑样本和特征对距离计算贡献程度的加权欧氏距离;为了确定最佳聚类数目的选取方法,依据改进的加权欧氏距离构造了加权簇内距离和簇间距离。最后,为克服传统K-means聚类目标函数仅考虑簇内的距离而未考虑簇间距离的缺陷,提出基于中位数的轮廓系数的目标函数,进而设计了DCFSK。实验结果表明,在10个基准测试函数上,IDCS的各项指标取得了较优的结果;相较于K-means、DBSCAN(Density-Based Spatial Clustering of Applications with Noise)等算法,在6个合成数据集与6个UCI数据集上,DCFSK的聚类效果最佳。展开更多
The first six Chebyshev polynomial coefficients (i.e., A00, A01, A10, A11, A02, A20) were derived from monthly mean geopotential height over East Asia for the period 1951-1983. Spectral analysis of these coefficients ...The first six Chebyshev polynomial coefficients (i.e., A00, A01, A10, A11, A02, A20) were derived from monthly mean geopotential height over East Asia for the period 1951-1983. Spectral analysis of these coefficients reveals relative maxima of power in the frequency bands of 200 months (- 16.7 years), 25 months (the quasi-biennial oscillation), 5-6 months, and 2-3 months. Cross-spectral characteristics between Chebyshev coefficients and the Southern Oscillation Index (SOI) were also explored. Coherence spectrum for the zonal and meridional circulation index (A01 and A 10) with the SOI was significant near 4 years, the QBO, and 2-3 months. Some physical explanations were offered for the spatial linkages (i.e., teleconnections) between the SO and atmospheric circulation anomalies overEast Asia.展开更多
An orthorhombic polycrystal is an orthorhombic aggregate of tiny crystallites. In this paper, we study the effect of the crystalline mean shape on the constitutive relation of the orthorhombic polycrystal. The crystal...An orthorhombic polycrystal is an orthorhombic aggregate of tiny crystallites. In this paper, we study the effect of the crystalline mean shape on the constitutive relation of the orthorhombic polycrystal. The crystalline mean shape and the crystalline orientation arrangement are described by the crystalline shape function （CSF） and the orientation distribution function （ODF）, respectively. The CSF and the ODF are expanded as an infinite series in terms of the Wigner D-functions. The expanded coefficients of the CSF and the ODF are called the shape coefficients s^lm0 and the texture coefficients c^lmn respectively. Assuming that Ceff in the constitutive relation depends on the shape coefficients s^lm0 and the texture coefficients c^lmn by the principle of material frame-indifference we derive an analytical expression for C^eff up to terms linear in s^lmo and c^lmn and the expression would be applicable to the polycrystal whose texture is weak and whose crystalline mean shape has weak anisotropy. C^cff contains six unspecified material constants （λ, μ, c, s1, s2, s3）, five shape coefficients （s^2 00, s^2 20, s^4 00, s^4 20, s^4 40）, and three texture coefficients （c^4 99,c^4 20, c^4 40）, The results based on the perturbation approach are used to determine the five material constants approximately. We also find that the shape coefficients 2 and a s^2mo and s^4m0 are all zero if the crystalline mean shape is a cuboid. Some examples are given to compare our computational results.展开更多
In order to study the temporal variations of correlations between two time series,a running correlation coefficient(RCC)could be used.An RCC is calculated for a given time window,and the window is then moved sequentia...In order to study the temporal variations of correlations between two time series,a running correlation coefficient(RCC)could be used.An RCC is calculated for a given time window,and the window is then moved sequentially through time.The current calculation method for RCCs is based on the general definition of the Pearson product-moment correlation coefficient,calculated with the data within the time window,which we call the local running correlation coefficient(LRCC).The LRCC is calculated via the two anomalies corresponding to the two local means,meanwhile,the local means also vary.It is cleared up that the LRCC reflects only the correlation between the two anomalies within the time window but fails to exhibit the contributions of the two varying means.To address this problem,two unchanged means obtained from all available data are adopted to calculate an RCC,which is called the synthetic running correlation coefficient(SRCC).When the anomaly variations are dominant,the two RCCs are similar.However,when the variations of the means are dominant,the difference between the two RCCs becomes obvious.The SRCC reflects the correlations of both the anomaly variations and the variations of the means.Therefore,the SRCCs from different time points are intercomparable.A criterion for the superiority of the RCC algorithm is that the average value of the RCC should be close to the global correlation coefficient calculated using all data.The SRCC always meets this criterion,while the LRCC sometimes fails.Therefore,the SRCC is better than the LRCC for running correlations.We suggest using the SRCC to calculate the RCCs.展开更多
Based on the review of present force coefficients estimation methods, a new method in the frequency domain, revised cross-spectrum estimation method, is presented in this paper. Some experiments on the wave-current fo...Based on the review of present force coefficients estimation methods, a new method in the frequency domain, revised cross-spectrum estimation method, is presented in this paper. Some experiments on the wave-current force on inclined cylinders are also described and the wave current force coefficients are estimated by the revised cross-spectrum estimation method. From the results, it is found that the wave and current directions have some regular effect on the coefficients. According to the results, some empirical formulas are obtained for converting the wave-current force coefficients on inclined cylinders into a unified coefficient. Comparisons show that the unified coefficients are in good agreement with other results.展开更多
针对如何快速实现地面观测获得的时序天文FITS(Flexible Image Transport System)图像的自动化分类和检验问题,提出了一种FITS图像自检验和自分类方法,该方法结合了K-means聚类算法及其思想,同时加入了一种基于皮尔逊相关系数的相似度...针对如何快速实现地面观测获得的时序天文FITS(Flexible Image Transport System)图像的自动化分类和检验问题,提出了一种FITS图像自检验和自分类方法,该方法结合了K-means聚类算法及其思想,同时加入了一种基于皮尔逊相关系数的相似度算法。通过比较该方法与基于有监督的VGG13分类网络和基于无监督的K-means聚类算法应用于真实的天文数据分类得到的错误数量,得出该方法的分类准确率达94%以上。该方法一方面检验出了历史数据中存在的错误情况,摆脱了对关键词IMAGETYP和观测日志的依赖,进一步规范和完善了历史存储的天文FITS数据;另一方面增强了分类的可靠性,提高了数据获取效率,降低了人工成本。展开更多
文摘K-means聚类算法随机确定初始聚类数目,而且原始数据集中含有大量的冗余特征会导致聚类时精度降低,而布谷鸟搜索(CS)算法存在收敛速度慢和局部搜索能力弱等问题,为此提出一种基于自适应布谷鸟优化特征选择的K-means聚类算法(DCFSK)。首先,为提升CS算法的搜索速度和精度,在莱维飞行阶段,设计了自适应步长因子;为调节CS算法全局搜索和局部搜索之间的平衡、加快CS算法的收敛,动态调整发现概率,进而提出改进的动态CS算法(IDCS),在IDCS的基础上构建了结合动态CS的特征选择算法(DCFS)。其次,为提升传统欧氏距离的计算精确度,设计同时考虑样本和特征对距离计算贡献程度的加权欧氏距离;为了确定最佳聚类数目的选取方法,依据改进的加权欧氏距离构造了加权簇内距离和簇间距离。最后,为克服传统K-means聚类目标函数仅考虑簇内的距离而未考虑簇间距离的缺陷,提出基于中位数的轮廓系数的目标函数,进而设计了DCFSK。实验结果表明,在10个基准测试函数上,IDCS的各项指标取得了较优的结果;相较于K-means、DBSCAN(Density-Based Spatial Clustering of Applications with Noise)等算法,在6个合成数据集与6个UCI数据集上,DCFSK的聚类效果最佳。
文摘The first six Chebyshev polynomial coefficients (i.e., A00, A01, A10, A11, A02, A20) were derived from monthly mean geopotential height over East Asia for the period 1951-1983. Spectral analysis of these coefficients reveals relative maxima of power in the frequency bands of 200 months (- 16.7 years), 25 months (the quasi-biennial oscillation), 5-6 months, and 2-3 months. Cross-spectral characteristics between Chebyshev coefficients and the Southern Oscillation Index (SOI) were also explored. Coherence spectrum for the zonal and meridional circulation index (A01 and A 10) with the SOI was significant near 4 years, the QBO, and 2-3 months. Some physical explanations were offered for the spatial linkages (i.e., teleconnections) between the SO and atmospheric circulation anomalies overEast Asia.
基金The project supported by the National Natural Science Foundation of China(10562004)the Oversea Returning Grant of China.
文摘An orthorhombic polycrystal is an orthorhombic aggregate of tiny crystallites. In this paper, we study the effect of the crystalline mean shape on the constitutive relation of the orthorhombic polycrystal. The crystalline mean shape and the crystalline orientation arrangement are described by the crystalline shape function （CSF） and the orientation distribution function （ODF）, respectively. The CSF and the ODF are expanded as an infinite series in terms of the Wigner D-functions. The expanded coefficients of the CSF and the ODF are called the shape coefficients s^lm0 and the texture coefficients c^lmn respectively. Assuming that Ceff in the constitutive relation depends on the shape coefficients s^lm0 and the texture coefficients c^lmn by the principle of material frame-indifference we derive an analytical expression for C^eff up to terms linear in s^lmo and c^lmn and the expression would be applicable to the polycrystal whose texture is weak and whose crystalline mean shape has weak anisotropy. C^cff contains six unspecified material constants （λ, μ, c, s1, s2, s3）, five shape coefficients （s^2 00, s^2 20, s^4 00, s^4 20, s^4 40）, and three texture coefficients （c^4 99,c^4 20, c^4 40）, The results based on the perturbation approach are used to determine the five material constants approximately. We also find that the shape coefficients 2 and a s^2mo and s^4m0 are all zero if the crystalline mean shape is a cuboid. Some examples are given to compare our computational results.
基金supported by the Key Program of the National Natural Science Foundation of China (No. 41330960)the Global Change Research Program of China (No. 2015CB953900)
文摘In order to study the temporal variations of correlations between two time series,a running correlation coefficient(RCC)could be used.An RCC is calculated for a given time window,and the window is then moved sequentially through time.The current calculation method for RCCs is based on the general definition of the Pearson product-moment correlation coefficient,calculated with the data within the time window,which we call the local running correlation coefficient(LRCC).The LRCC is calculated via the two anomalies corresponding to the two local means,meanwhile,the local means also vary.It is cleared up that the LRCC reflects only the correlation between the two anomalies within the time window but fails to exhibit the contributions of the two varying means.To address this problem,two unchanged means obtained from all available data are adopted to calculate an RCC,which is called the synthetic running correlation coefficient(SRCC).When the anomaly variations are dominant,the two RCCs are similar.However,when the variations of the means are dominant,the difference between the two RCCs becomes obvious.The SRCC reflects the correlations of both the anomaly variations and the variations of the means.Therefore,the SRCCs from different time points are intercomparable.A criterion for the superiority of the RCC algorithm is that the average value of the RCC should be close to the global correlation coefficient calculated using all data.The SRCC always meets this criterion,while the LRCC sometimes fails.Therefore,the SRCC is better than the LRCC for running correlations.We suggest using the SRCC to calculate the RCCs.
文摘Based on the review of present force coefficients estimation methods, a new method in the frequency domain, revised cross-spectrum estimation method, is presented in this paper. Some experiments on the wave-current force on inclined cylinders are also described and the wave current force coefficients are estimated by the revised cross-spectrum estimation method. From the results, it is found that the wave and current directions have some regular effect on the coefficients. According to the results, some empirical formulas are obtained for converting the wave-current force coefficients on inclined cylinders into a unified coefficient. Comparisons show that the unified coefficients are in good agreement with other results.
文摘针对如何快速实现地面观测获得的时序天文FITS(Flexible Image Transport System)图像的自动化分类和检验问题,提出了一种FITS图像自检验和自分类方法,该方法结合了K-means聚类算法及其思想,同时加入了一种基于皮尔逊相关系数的相似度算法。通过比较该方法与基于有监督的VGG13分类网络和基于无监督的K-means聚类算法应用于真实的天文数据分类得到的错误数量,得出该方法的分类准确率达94%以上。该方法一方面检验出了历史数据中存在的错误情况,摆脱了对关键词IMAGETYP和观测日志的依赖,进一步规范和完善了历史存储的天文FITS数据;另一方面增强了分类的可靠性,提高了数据获取效率,降低了人工成本。
