标记分布学习是近年来提出的一种新的机器学习范式,它能很好地解决某些标记多义性的问题。现有的标记分布学习算法均利用条件概率建立参数模型,但未能充分利用特征和标记间的联系。本文考虑到特征相似的样本所对应的标记分布也应当相似...标记分布学习是近年来提出的一种新的机器学习范式,它能很好地解决某些标记多义性的问题。现有的标记分布学习算法均利用条件概率建立参数模型,但未能充分利用特征和标记间的联系。本文考虑到特征相似的样本所对应的标记分布也应当相似,利用原型聚类的k均值算法(k-means),将训练集的样本进行聚类,提出基于kmeans算法的标记分布学习(label distribution learning based on k-means algorithm,LDLKM)。首先通过聚类算法kmeans求得每一个簇的均值向量,然后分别求得对应标记分布的均值向量。最后将测试集和训练集的均值向量间的距离作为权重,应用到对测试集标记分布的预测上。在6个公开的数据集上进行实验,并与3种已有的标记分布学习算法在5种评价指标上进行比较,实验结果表明提出的LDLKM算法是有效的。展开更多
This study compares the climatology and long-term trend of northern winter stratospheric residual mean meridional circulation(RMMC), as well as its responses to El Ni?o-Southern Oscillation(ENSO), stratospheric Quasi ...This study compares the climatology and long-term trend of northern winter stratospheric residual mean meridional circulation(RMMC), as well as its responses to El Ni?o-Southern Oscillation(ENSO), stratospheric Quasi Biennial Oscillation(QBO), and solar cycle in ten reanalyses and a stratosphere-resolving model, CESM1-WACCM. The RMMC is a large-scale meridional circulation cell in the stratosphere, usually referred to as the estimate of the Brewer Dobson circulation(BDC). The distribution of the BDC is generally consistent among multiple reanalyses except that the NOAA twentieth century reanalysis(20RC) largely underestimates it. Most reanalyses(except ERA40 and ERA-Interim) show a strengthening trend for the BDC during 1979–2010. All reanalyses and CESM1-WACCM consistently reveal that the deep branch of the BDC is significantly enhanced in El Ni?o winters as more waves from the troposphere dissipate in the stratospheric polar vortex region. A secondary circulation cell is coupled to the temperature anomalies below the QBO easterly center at 50 hPa with tropical upwelling/cooling and midlatitude downwelling/warming, and similar secondary circulation cells also appear between 50–10 hPa and above 10 hPa to balance the temperature anomalies. The direct BDC response to QBO in the upper stratosphere creates a barrier near 30°N to prevent waves from propagating to midlatitudes, contributing to the weakening of the polar vortex. The shallow branch of the BDC in the lower stratosphere is intensified during solar minima, and the downwelling warms the Arctic lower stratosphere. The stratospheric responses to QBO and solar cycle in most reanalyses are generally consistent except in the two 20 CRs.展开更多
文摘标记分布学习是近年来提出的一种新的机器学习范式,它能很好地解决某些标记多义性的问题。现有的标记分布学习算法均利用条件概率建立参数模型,但未能充分利用特征和标记间的联系。本文考虑到特征相似的样本所对应的标记分布也应当相似,利用原型聚类的k均值算法(k-means),将训练集的样本进行聚类,提出基于kmeans算法的标记分布学习(label distribution learning based on k-means algorithm,LDLKM)。首先通过聚类算法kmeans求得每一个簇的均值向量,然后分别求得对应标记分布的均值向量。最后将测试集和训练集的均值向量间的距离作为权重,应用到对测试集标记分布的预测上。在6个公开的数据集上进行实验,并与3种已有的标记分布学习算法在5种评价指标上进行比较,实验结果表明提出的LDLKM算法是有效的。
基金supported by grants from the National Natural Science Foundation of China(41705024,41875048)the National Key R&D Program of China(2016YFA0602104)+1 种基金the Planning and Budgeting Committee of the Council for Higher Education in Israelthe Startup Foundation for Introducing Talent of NUIST(2016r060)
文摘This study compares the climatology and long-term trend of northern winter stratospheric residual mean meridional circulation(RMMC), as well as its responses to El Ni?o-Southern Oscillation(ENSO), stratospheric Quasi Biennial Oscillation(QBO), and solar cycle in ten reanalyses and a stratosphere-resolving model, CESM1-WACCM. The RMMC is a large-scale meridional circulation cell in the stratosphere, usually referred to as the estimate of the Brewer Dobson circulation(BDC). The distribution of the BDC is generally consistent among multiple reanalyses except that the NOAA twentieth century reanalysis(20RC) largely underestimates it. Most reanalyses(except ERA40 and ERA-Interim) show a strengthening trend for the BDC during 1979–2010. All reanalyses and CESM1-WACCM consistently reveal that the deep branch of the BDC is significantly enhanced in El Ni?o winters as more waves from the troposphere dissipate in the stratospheric polar vortex region. A secondary circulation cell is coupled to the temperature anomalies below the QBO easterly center at 50 hPa with tropical upwelling/cooling and midlatitude downwelling/warming, and similar secondary circulation cells also appear between 50–10 hPa and above 10 hPa to balance the temperature anomalies. The direct BDC response to QBO in the upper stratosphere creates a barrier near 30°N to prevent waves from propagating to midlatitudes, contributing to the weakening of the polar vortex. The shallow branch of the BDC in the lower stratosphere is intensified during solar minima, and the downwelling warms the Arctic lower stratosphere. The stratospheric responses to QBO and solar cycle in most reanalyses are generally consistent except in the two 20 CRs.
