随着智能移动设备的快速普及,人们对基于位置的社交网络服务的依赖性越来越高.但是,由于数据采集成本昂贵以及现有数据采集技术的缺陷,基于小样本数据挖掘的兴趣点(point of interest, POI)定位已经成为了一种挑战.尽管已经有一些POI定...随着智能移动设备的快速普及,人们对基于位置的社交网络服务的依赖性越来越高.但是,由于数据采集成本昂贵以及现有数据采集技术的缺陷,基于小样本数据挖掘的兴趣点(point of interest, POI)定位已经成为了一种挑战.尽管已经有一些POI定位方面的研究,但是现有的方法不能解决正样本数据不足的问题.提出一种基于PU与生成对抗网络(positive and unlabeled generative adversarial network, puGAN)的模型,采用PU学习和生成对抗网络相结合的方式挖掘数据的隐藏特征,生成伪正样本弥补数据不足的问题,并校正无标签样本数据的分布,从而训练出有效的POI判别模型.通过分析ROC曲线以及训练误差和测试误差在迭代过程中的变化和关系来比较不同模型在实验场景下的效果.结果表明,puGAN模型可以有效解决数据样本不足的问题,进而提高POI定位的准确性.展开更多
A total of 137 internal solitary waves （ISWs） are captured during a field experiment conducted in the deep basin west of the Luzon Strait （LS） from March to August, 2010. Mooring observations reveal that a fully d...A total of 137 internal solitary waves （ISWs） are captured during a field experiment conducted in the deep basin west of the Luzon Strait （LS） from March to August, 2010. Mooring observations reveal that a fully developed ISW owns a maximum westward velocity of more than 1.8 m/s and an amplitude of about 200 m. The ISWs in the South China Sea （SCS） are most active in July, which may be due to the strong stratification in summer. Most of the ISW episodes are detected around and after the 1st or 15th lunar day, indicating that the ISW in the SCS is triggered by astronomic tides. Half part of the observed ISWs were detected around 19：00 local time, which can be explained by the fact that type-a ISWs emerged in the evening at roughly the same time each day. The propagation direction of the ISWs and the astronomic tides in the LS show that the area south of the Batan Island is probably the main source region of the type-a ISWs, while the area south of Itbayat Island and south of the Batan Island is likely the main source region of the type-b ISWs observed at the mooring. Moreover, for the resonance of semidiurnal internal tides emitting from the double ridges in the LS, the underwater ridge south of the Itbayat Island and south of the Batan Island is believed to favor the generation of the energetic ISWs.展开更多
The Argo data are used to calculate eddy（turbulence） heat transport（EHT） in the global ocean and analyze its horizontal distribution and vertical structure.We calculate the EHT by averaging all the v ′,T ′ profi...The Argo data are used to calculate eddy（turbulence） heat transport（EHT） in the global ocean and analyze its horizontal distribution and vertical structure.We calculate the EHT by averaging all the v ′,T ′ profiles within each 2 ×2 bin.The velocity and temperature anomalies are obtained by removing their climatological values from the Argo ＂instantaneous＂ values respectively.Through the Student＇s t-test and an error evaluation,we obtained a total of 87% Argo bins with significant depth-integrated EHTs（D-EHTs）.The results reveal a positive-and-negative alternating D-EHT pattern along the western boundary currents（WBC） and Antarctic Circumpolar Current（ACC）.The zonally-integrated D-EHT（ZI-EHT） of the global ocean reaches 0.12 PW in the northern WBC band and –0.38 PW in the ACC band respectively.The strong ZI-EHT across the ACC in the global ocean is mainly caused by the southern Indian Ocean.The ZI-EHT in the above two bands accounts for a large portion of the total oceanic heat transport,which may play an important role in regulating the climate.The analysis of vertical structures of the EHT along the 35 N and 45 S section reveals that the oscillating EHT pattern can reach deep in the northern WBC regions and the Agulhas Return Current（ARC） region.It also shows that the strong EHT could reach 600 m in the WBC regions and 1 000 m in the ARC region,with the maximum mainly located between 100 and 400 m depth.The results would provide useful information for improving the parameterization scheme in models.展开更多
文摘随着智能移动设备的快速普及,人们对基于位置的社交网络服务的依赖性越来越高.但是,由于数据采集成本昂贵以及现有数据采集技术的缺陷,基于小样本数据挖掘的兴趣点(point of interest, POI)定位已经成为了一种挑战.尽管已经有一些POI定位方面的研究,但是现有的方法不能解决正样本数据不足的问题.提出一种基于PU与生成对抗网络(positive and unlabeled generative adversarial network, puGAN)的模型,采用PU学习和生成对抗网络相结合的方式挖掘数据的隐藏特征,生成伪正样本弥补数据不足的问题,并校正无标签样本数据的分布,从而训练出有效的POI判别模型.通过分析ROC曲线以及训练误差和测试误差在迭代过程中的变化和关系来比较不同模型在实验场景下的效果.结果表明,puGAN模型可以有效解决数据样本不足的问题,进而提高POI定位的准确性.
基金The Program for New Century Excellent Talents in Universities of China under contract No.NCET-10-0764the National High Tech-nology Research and Development(863)Program of China under contract No.2013AA09A502the National Natural Science Foundation of China under contract Nos 41176010,91028008,40890153,41176008 and 40776005
文摘A total of 137 internal solitary waves （ISWs） are captured during a field experiment conducted in the deep basin west of the Luzon Strait （LS） from March to August, 2010. Mooring observations reveal that a fully developed ISW owns a maximum westward velocity of more than 1.8 m/s and an amplitude of about 200 m. The ISWs in the South China Sea （SCS） are most active in July, which may be due to the strong stratification in summer. Most of the ISW episodes are detected around and after the 1st or 15th lunar day, indicating that the ISW in the SCS is triggered by astronomic tides. Half part of the observed ISWs were detected around 19：00 local time, which can be explained by the fact that type-a ISWs emerged in the evening at roughly the same time each day. The propagation direction of the ISWs and the astronomic tides in the LS show that the area south of the Batan Island is probably the main source region of the type-a ISWs, while the area south of Itbayat Island and south of the Batan Island is likely the main source region of the type-b ISWs observed at the mooring. Moreover, for the resonance of semidiurnal internal tides emitting from the double ridges in the LS, the underwater ridge south of the Itbayat Island and south of the Batan Island is believed to favor the generation of the energetic ISWs.
基金The Major Program of the National Natural Science Foundation of China under contact No.40890153The National High Tech-nology Research and Development Program of China(863 Program)under contact No.2008AA09A402
文摘The Argo data are used to calculate eddy（turbulence） heat transport（EHT） in the global ocean and analyze its horizontal distribution and vertical structure.We calculate the EHT by averaging all the v ′,T ′ profiles within each 2 ×2 bin.The velocity and temperature anomalies are obtained by removing their climatological values from the Argo ＂instantaneous＂ values respectively.Through the Student＇s t-test and an error evaluation,we obtained a total of 87% Argo bins with significant depth-integrated EHTs（D-EHTs）.The results reveal a positive-and-negative alternating D-EHT pattern along the western boundary currents（WBC） and Antarctic Circumpolar Current（ACC）.The zonally-integrated D-EHT（ZI-EHT） of the global ocean reaches 0.12 PW in the northern WBC band and –0.38 PW in the ACC band respectively.The strong ZI-EHT across the ACC in the global ocean is mainly caused by the southern Indian Ocean.The ZI-EHT in the above two bands accounts for a large portion of the total oceanic heat transport,which may play an important role in regulating the climate.The analysis of vertical structures of the EHT along the 35 N and 45 S section reveals that the oscillating EHT pattern can reach deep in the northern WBC regions and the Agulhas Return Current（ARC） region.It also shows that the strong EHT could reach 600 m in the WBC regions and 1 000 m in the ARC region,with the maximum mainly located between 100 and 400 m depth.The results would provide useful information for improving the parameterization scheme in models.
