We proposed a method using latent regression Bayesian network (LRBN) toextract the shared speech feature for the input of end-to-end speech recognition model.The structure of LRBN is compact and its parameter learning...We proposed a method using latent regression Bayesian network (LRBN) toextract the shared speech feature for the input of end-to-end speech recognition model.The structure of LRBN is compact and its parameter learning is fast. Compared withConvolutional Neural Network, it has a simpler and understood structure and lessparameters to learn. Experimental results show that the advantage of hybridLRBN/Bidirectional Long Short-Term Memory-Connectionist Temporal Classificationarchitecture for Tibetan multi-dialect speech recognition, and demonstrate the LRBN ishelpful to differentiate among multiple language speech sets.展开更多
The abundance of social networking platforms has increased the frequency and the availability for which individuals communicate with one another. The feasibility and accessibility to go online to find sexual partners ...The abundance of social networking platforms has increased the frequency and the availability for which individuals communicate with one another. The feasibility and accessibility to go online to find sexual partners pose opportunity for contracting sexually transmitted infections (STI) in the absence of safe sexual practices. Low condom use has been reported among young adults who seek sexual partners online. African American young adults have some of the highest rates of infection for certain STIs. In order to mitigate the incidence and prevalence of STIs in at-risk populations, sexually active young adults must use condoms consistently and correctly during sexual activities. The present study sought to uncover the heterogeneity within African American young adults regarding their online networking utilization, STI knowledge, and sexual risk behavior. African American young adults (N = 236), ages 18 - 23, completed private online survey administration. Using latent class analysis, three classes were identified: Social Network Communicators (43%;N = 101), Social Networking Daters (36%;N = 83), and Media Sharers (21%;N = 52). Social Networking Daters exhibited the highest probability of using online dating sites daily, low STI knowledge, and a zero probability of consistent condom use. All three groups exhibited relatively low STI knowledge. Furthermore, having a history of STI increased the likelihood of being classified into the Social Networking Daters class relative to the other classes. Findings highlight the need to capitalize upon online platforms for African American young adults who utilize online dating sites and other online environments.展开更多
在全球气候变化背景下,洪涝灾害已成为威胁人群生命财产安全的世界性问题。目前,关于洪涝灾害的研究已较为丰富,需进行系统性梳理和总结。为探究国内外洪涝灾害的研究现状与趋势,采用文献计量的分析方法,对2003—2022年中国知网(China N...在全球气候变化背景下,洪涝灾害已成为威胁人群生命财产安全的世界性问题。目前,关于洪涝灾害的研究已较为丰富,需进行系统性梳理和总结。为探究国内外洪涝灾害的研究现状与趋势,采用文献计量的分析方法,对2003—2022年中国知网(China National Knowledge Infrastructure,CNKI)和WoS(Web of Science)数据库中主题为洪涝灾害的中英文文献进行文本分析。从关键词共现模式、研究主题的时间聚类及空间分布三个角度,探索洪涝灾害相关研究的主题演进模式和地理分布特征,并总结未来发展方向。结果表明:(1)在研究关键词共现模式方面,中文研究更加关心洪涝灾害区域影响及管理策略等灾害的后续影响方面,英文研究则更倾向于探究洪涝灾害的成因,从气候变化与孕灾环境角度分析洪涝灾害。中英文研究均强调新兴技术在洪涝灾害研究中的应用。(2)在研究主题时间聚类方面,中文研究在21世纪初期追随英文研究的关注热点,并在2015年前后逐渐形成具有中国本土化特色的研究框架,强调使用多源数据和多种算法模型进行定量化分析,主要着眼于城市化发展对洪涝灾害的多时段、多尺度干预效应。(3)在研究主题空间分布方面,沿海经济发达地区受到研究者更多的关注,表明洪涝灾害研究的主题与数量受到研究区域的地理位置、气候条件和社会经济发展水平的影响。展开更多
人脸年龄合成(face age synthesis,FAS)的目标是根据源人脸图像合成指定年龄人脸图像,同时保留人脸的个人特征和身份信息.针对年龄变换时无关特征容易改变和产生伪影鬼影的问题,提出一种基于生成对抗网络的人脸年龄渐进合成算法.采用基...人脸年龄合成(face age synthesis,FAS)的目标是根据源人脸图像合成指定年龄人脸图像,同时保留人脸的个人特征和身份信息.针对年龄变换时无关特征容易改变和产生伪影鬼影的问题,提出一种基于生成对抗网络的人脸年龄渐进合成算法.采用基于门控循环单元的年龄编辑模块自适应地过滤或加入特征,并使用属性解耦模块在潜在空间进行对抗学习,通过生成器和判别器的对抗策略保证了真实自然的人脸合成,使用年龄分类约束拟合特定年龄分布,为了保证年龄无关属性的保留,还在生成对抗网络中引入了重建学习.在跨年龄名人数据集(cross-age celebrity dataset,CACD)下的实验结果表明,对比其他基于条件生成对抗网络的算法,提出的算法生成的人脸图像伪影失真有所减少,年龄显著性增强,具有较好的年龄准确性和较高的身份一致性.展开更多
为了更好地对社会网络提供动态分析,提出了一种新的基于事件的动态社会网络分析算法(Dynamic Social Network Analysis Algorithm Based on Events,DSNE)。该算法基于隐空间和两阶段聚类方法充分利用实体和事件的动态信息,能够很好地确...为了更好地对社会网络提供动态分析,提出了一种新的基于事件的动态社会网络分析算法(Dynamic Social Network Analysis Algorithm Based on Events,DSNE)。该算法基于隐空间和两阶段聚类方法充分利用实体和事件的动态信息,能够很好地确定每个簇的核心节点,并根据时间步的变化观察到节点位置的变化趋势。实验结果表明了算法的可行性、有效性和准确性。展开更多
文摘We proposed a method using latent regression Bayesian network (LRBN) toextract the shared speech feature for the input of end-to-end speech recognition model.The structure of LRBN is compact and its parameter learning is fast. Compared withConvolutional Neural Network, it has a simpler and understood structure and lessparameters to learn. Experimental results show that the advantage of hybridLRBN/Bidirectional Long Short-Term Memory-Connectionist Temporal Classificationarchitecture for Tibetan multi-dialect speech recognition, and demonstrate the LRBN ishelpful to differentiate among multiple language speech sets.
文摘The abundance of social networking platforms has increased the frequency and the availability for which individuals communicate with one another. The feasibility and accessibility to go online to find sexual partners pose opportunity for contracting sexually transmitted infections (STI) in the absence of safe sexual practices. Low condom use has been reported among young adults who seek sexual partners online. African American young adults have some of the highest rates of infection for certain STIs. In order to mitigate the incidence and prevalence of STIs in at-risk populations, sexually active young adults must use condoms consistently and correctly during sexual activities. The present study sought to uncover the heterogeneity within African American young adults regarding their online networking utilization, STI knowledge, and sexual risk behavior. African American young adults (N = 236), ages 18 - 23, completed private online survey administration. Using latent class analysis, three classes were identified: Social Network Communicators (43%;N = 101), Social Networking Daters (36%;N = 83), and Media Sharers (21%;N = 52). Social Networking Daters exhibited the highest probability of using online dating sites daily, low STI knowledge, and a zero probability of consistent condom use. All three groups exhibited relatively low STI knowledge. Furthermore, having a history of STI increased the likelihood of being classified into the Social Networking Daters class relative to the other classes. Findings highlight the need to capitalize upon online platforms for African American young adults who utilize online dating sites and other online environments.
文摘在全球气候变化背景下,洪涝灾害已成为威胁人群生命财产安全的世界性问题。目前,关于洪涝灾害的研究已较为丰富,需进行系统性梳理和总结。为探究国内外洪涝灾害的研究现状与趋势,采用文献计量的分析方法,对2003—2022年中国知网(China National Knowledge Infrastructure,CNKI)和WoS(Web of Science)数据库中主题为洪涝灾害的中英文文献进行文本分析。从关键词共现模式、研究主题的时间聚类及空间分布三个角度,探索洪涝灾害相关研究的主题演进模式和地理分布特征,并总结未来发展方向。结果表明:(1)在研究关键词共现模式方面,中文研究更加关心洪涝灾害区域影响及管理策略等灾害的后续影响方面,英文研究则更倾向于探究洪涝灾害的成因,从气候变化与孕灾环境角度分析洪涝灾害。中英文研究均强调新兴技术在洪涝灾害研究中的应用。(2)在研究主题时间聚类方面,中文研究在21世纪初期追随英文研究的关注热点,并在2015年前后逐渐形成具有中国本土化特色的研究框架,强调使用多源数据和多种算法模型进行定量化分析,主要着眼于城市化发展对洪涝灾害的多时段、多尺度干预效应。(3)在研究主题空间分布方面,沿海经济发达地区受到研究者更多的关注,表明洪涝灾害研究的主题与数量受到研究区域的地理位置、气候条件和社会经济发展水平的影响。
文摘人脸年龄合成(face age synthesis,FAS)的目标是根据源人脸图像合成指定年龄人脸图像,同时保留人脸的个人特征和身份信息.针对年龄变换时无关特征容易改变和产生伪影鬼影的问题,提出一种基于生成对抗网络的人脸年龄渐进合成算法.采用基于门控循环单元的年龄编辑模块自适应地过滤或加入特征,并使用属性解耦模块在潜在空间进行对抗学习,通过生成器和判别器的对抗策略保证了真实自然的人脸合成,使用年龄分类约束拟合特定年龄分布,为了保证年龄无关属性的保留,还在生成对抗网络中引入了重建学习.在跨年龄名人数据集(cross-age celebrity dataset,CACD)下的实验结果表明,对比其他基于条件生成对抗网络的算法,提出的算法生成的人脸图像伪影失真有所减少,年龄显著性增强,具有较好的年龄准确性和较高的身份一致性.
文摘为了更好地对社会网络提供动态分析,提出了一种新的基于事件的动态社会网络分析算法(Dynamic Social Network Analysis Algorithm Based on Events,DSNE)。该算法基于隐空间和两阶段聚类方法充分利用实体和事件的动态信息,能够很好地确定每个簇的核心节点,并根据时间步的变化观察到节点位置的变化趋势。实验结果表明了算法的可行性、有效性和准确性。
