由于卷积滤波尺寸等限制,U-net无法学习到心电(Electrocardiographic,ECG)信号的长时序关联性以及标签间的相关性。对此提出一种基于U-net-BiLSTM-CRF的心律失常多目标检测方法,可同时输出目标心拍所属类型和位置信息。使用U-net学习融...由于卷积滤波尺寸等限制,U-net无法学习到心电(Electrocardiographic,ECG)信号的长时序关联性以及标签间的相关性。对此提出一种基于U-net-BiLSTM-CRF的心律失常多目标检测方法,可同时输出目标心拍所属类型和位置信息。使用U-net学习融合特征,再将其输入到双向长短时记忆网络(Bi-directional Long Short-Term Memory,BiLSTM)中学习长时序依赖特征,最后使用条件随机场(Conditional Random Field,CRF)对标签间的关系建模,优化分类结果。依据ANSI/AAMI EC57:2012的心搏分类标准,对MIT-BIH心律失常数据库中共85609个心拍记录进行划分,在划分后数据集上的实验结果表明,该方法对心拍分类的准确率达99.11%,特异性为99.76%,灵敏度为97.21%,优于传统U-net在MIT-BIH心律失常数据库上的分类性能。展开更多
Atlantic meridional overturning circulation(AMOC)plays an important role in transporting heat meridionally in the Earth’s climate system and is also a key metrical tool to verify oceanic general circulation models.Tw...Atlantic meridional overturning circulation(AMOC)plays an important role in transporting heat meridionally in the Earth’s climate system and is also a key metrical tool to verify oceanic general circulation models.Two OMIP(Ocean Model Intercomparison Project phase 1 and 2)simulations with LICOM3(version 3 of the LASG/IAP Climate System Ocean Model)developed at the State Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics(LASG),Institute of Atmospheric Physics(IAP),are compared in this study.Both simulations well reproduce the fundamental characteristics of the AMOC,but the OMIP1 simulation shows a significantly stronger AMOC than the OMIP2 simulation.Because the LICOM3 configurations are identical between these two experiments,any differences between them must be attributed to the surface forcing data.Further analysis suggests that sea surface salinity(SSS)differences should be mainly responsible for the enhanced AMOC in the OMIP1 simulation,but sea surface temperature(SST)also play an unignorable role in modulating AMOC.In the North Atlantic,where deep convection occurs,the SSS in OMIP1 is more saline than that in OMIP1.We find that in the major region of deep convection,the change of SSS has more significant effect on density than the change of SST.As a result,the SSS was more saline than that in OMIP2,leading to stronger deep convection and subsequently intensify the AMOC.We conduct a series of numerical experiments with LICOM3,and the results confirmed that the changes in SSS have more significant effect on the strength of AMOC than the changes in SST.展开更多
针对油气领域知识图谱构建过程中命名实体识别使用传统方法存在实体特征信息提取不准确、识别效率低的问题,提出了一种基于BERT-BiLSTM-CRF模型的命名实体识别研究方法。该方法首先利用BERT(bidirectional encoder representations from...针对油气领域知识图谱构建过程中命名实体识别使用传统方法存在实体特征信息提取不准确、识别效率低的问题,提出了一种基于BERT-BiLSTM-CRF模型的命名实体识别研究方法。该方法首先利用BERT(bidirectional encoder representations from transformers)预训练模型得到输入序列语义的词向量;然后将训练后的词向量输入双向长短期记忆网络(bi-directional long short-term memory,BiLSTM)模型进一步获取上下文特征;最后根据条件随机场(conditional random fields,CRF)的标注规则和序列解码能力输出最大概率序列标注结果,构建油气领域命名实体识别模型框架。将BERT-BiLSTM-CRF模型与其他2种命名实体识别模型(BiLSTM-CRF、BiLSTM-Attention-CRF)在包括3万多条文本语料数据、4类实体的自建数据集上进行了对比实验。实验结果表明,BERT-BiLSTM-CRF模型的准确率(P)、召回率(R)和F_(1)值分别达到91.3%、94.5%和92.9%,实体识别效果优于其他2种模型。展开更多
The West African Monsoon (WAM) is characterized by strong decadal and multi-decadal variability and the impacts can be catastrophic for the local populations. One of the factors put forward to explain this variability...The West African Monsoon (WAM) is characterized by strong decadal and multi-decadal variability and the impacts can be catastrophic for the local populations. One of the factors put forward to explain this variability involves the role of atmospheric dynamics, linked in particular to the Saharan Heat Low (SHL). This article addresses this question by comparing the sets of preindustrial control and historical simulation data from climate models carried out in the framework of the CMIP5 project and observations data over the 20<sup>th</sup> century. Through multivariate statistical analyses, it was established that decadal modes of ocean variability and decadal variability of Saharan atmospheric dynamics significantly influence decadal variability of monsoon precipitation. These results also suggest the existence of external anthropogenic forcing, which is superimposed on the decadal natural variability inducing an intensification of the signal in the historical simulations compared to preindustrial control simulations. We have also shown that decadal rainfall variability in the Sahel, once the influence of oceanic modes has been eliminated, appears to be driven mainly by the activity of the Arabian Heat Low (AHL) in the central Sahel, and by the structure of the meridional temperature gradient over the inter-tropical Atlantic in the western Sahel.展开更多
文摘由于卷积滤波尺寸等限制,U-net无法学习到心电(Electrocardiographic,ECG)信号的长时序关联性以及标签间的相关性。对此提出一种基于U-net-BiLSTM-CRF的心律失常多目标检测方法,可同时输出目标心拍所属类型和位置信息。使用U-net学习融合特征,再将其输入到双向长短时记忆网络(Bi-directional Long Short-Term Memory,BiLSTM)中学习长时序依赖特征,最后使用条件随机场(Conditional Random Field,CRF)对标签间的关系建模,优化分类结果。依据ANSI/AAMI EC57:2012的心搏分类标准,对MIT-BIH心律失常数据库中共85609个心拍记录进行划分,在划分后数据集上的实验结果表明,该方法对心拍分类的准确率达99.11%,特异性为99.76%,灵敏度为97.21%,优于传统U-net在MIT-BIH心律失常数据库上的分类性能。
基金Supported by the Strategic Priority Research Program of the Chinese Academy of Sciences(No.XDA19060102)the National Natural Science Foundation of China(Nos.91958201,42130608)the Strategic Priority Research Program of the Chinese Academy of Sciences(No.XDB42000000)。
文摘Atlantic meridional overturning circulation(AMOC)plays an important role in transporting heat meridionally in the Earth’s climate system and is also a key metrical tool to verify oceanic general circulation models.Two OMIP(Ocean Model Intercomparison Project phase 1 and 2)simulations with LICOM3(version 3 of the LASG/IAP Climate System Ocean Model)developed at the State Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics(LASG),Institute of Atmospheric Physics(IAP),are compared in this study.Both simulations well reproduce the fundamental characteristics of the AMOC,but the OMIP1 simulation shows a significantly stronger AMOC than the OMIP2 simulation.Because the LICOM3 configurations are identical between these two experiments,any differences between them must be attributed to the surface forcing data.Further analysis suggests that sea surface salinity(SSS)differences should be mainly responsible for the enhanced AMOC in the OMIP1 simulation,but sea surface temperature(SST)also play an unignorable role in modulating AMOC.In the North Atlantic,where deep convection occurs,the SSS in OMIP1 is more saline than that in OMIP1.We find that in the major region of deep convection,the change of SSS has more significant effect on density than the change of SST.As a result,the SSS was more saline than that in OMIP2,leading to stronger deep convection and subsequently intensify the AMOC.We conduct a series of numerical experiments with LICOM3,and the results confirmed that the changes in SSS have more significant effect on the strength of AMOC than the changes in SST.
文摘针对油气领域知识图谱构建过程中命名实体识别使用传统方法存在实体特征信息提取不准确、识别效率低的问题,提出了一种基于BERT-BiLSTM-CRF模型的命名实体识别研究方法。该方法首先利用BERT(bidirectional encoder representations from transformers)预训练模型得到输入序列语义的词向量;然后将训练后的词向量输入双向长短期记忆网络(bi-directional long short-term memory,BiLSTM)模型进一步获取上下文特征;最后根据条件随机场(conditional random fields,CRF)的标注规则和序列解码能力输出最大概率序列标注结果,构建油气领域命名实体识别模型框架。将BERT-BiLSTM-CRF模型与其他2种命名实体识别模型(BiLSTM-CRF、BiLSTM-Attention-CRF)在包括3万多条文本语料数据、4类实体的自建数据集上进行了对比实验。实验结果表明,BERT-BiLSTM-CRF模型的准确率(P)、召回率(R)和F_(1)值分别达到91.3%、94.5%和92.9%,实体识别效果优于其他2种模型。
文摘The West African Monsoon (WAM) is characterized by strong decadal and multi-decadal variability and the impacts can be catastrophic for the local populations. One of the factors put forward to explain this variability involves the role of atmospheric dynamics, linked in particular to the Saharan Heat Low (SHL). This article addresses this question by comparing the sets of preindustrial control and historical simulation data from climate models carried out in the framework of the CMIP5 project and observations data over the 20<sup>th</sup> century. Through multivariate statistical analyses, it was established that decadal modes of ocean variability and decadal variability of Saharan atmospheric dynamics significantly influence decadal variability of monsoon precipitation. These results also suggest the existence of external anthropogenic forcing, which is superimposed on the decadal natural variability inducing an intensification of the signal in the historical simulations compared to preindustrial control simulations. We have also shown that decadal rainfall variability in the Sahel, once the influence of oceanic modes has been eliminated, appears to be driven mainly by the activity of the Arabian Heat Low (AHL) in the central Sahel, and by the structure of the meridional temperature gradient over the inter-tropical Atlantic in the western Sahel.