The phasing out of protective measures by governments and public health agencies, despite continued seriousness of the coronavirus pandemic, leaves individuals who are concerned for their health with two basic options...The phasing out of protective measures by governments and public health agencies, despite continued seriousness of the coronavirus pandemic, leaves individuals who are concerned for their health with two basic options over which they have control: 1) minimize risk of infection by being vaccinated and by wearing a face mask when appropriate, and 2) minimize risk of transmission upon infection by self-isolating. For the latter to be effective, it is essential to have an accurate sense of the probability of infectivity as a function of time following the onset of symptoms. Epidemiological considerations suggest that the period of infectivity follows a lognormal distribution. This proposition is tested empirically by construction of the lognormal probability density function and cumulative distribution function based on quantiles of infectivity reported by several independent investigations. A comprehensive examination of a prototypical ideal clinical study, based on general statistical principles (the Principle of Maximum Entropy and the Central Limit Theorem) reveals that the probability of infectivity is a lognormal random variable. Subsequent evolution of new variants may change the parameters of the distribution, which can be updated by the methods in this paper, but the form of the probability function is expected to remain lognormal as this is the most probable distribution consistent with mathematical requirements and available information.展开更多
大规模人脸聚类不仅要求高效的人脸特征,还要求聚类算法在保持高准确率的同时耗时短.本文通过构建卷积神经网络高效提取人脸特征,并采用经典K-means算法和现阶段新颖的CFSFDP(Clustering by Fast Search and Find of Density Peaks)算...大规模人脸聚类不仅要求高效的人脸特征,还要求聚类算法在保持高准确率的同时耗时短.本文通过构建卷积神经网络高效提取人脸特征,并采用经典K-means算法和现阶段新颖的CFSFDP(Clustering by Fast Search and Find of Density Peaks)算法进行大规模人脸聚类.实验在聚类数目递增的情况下进行,并通过随机指标(Rand Index,RI)、信息熵、F1-measure和混淆矩阵可视化来综合评估聚类的质量.结果表明,在大规模人脸聚类的情况下,卷积神经网络特征融合K-means的人脸聚类算法速度和准确率均优于CFSFDP算法.这一结论对大规模人脸聚类的实际应用具有重要的指导意义.展开更多
文摘The phasing out of protective measures by governments and public health agencies, despite continued seriousness of the coronavirus pandemic, leaves individuals who are concerned for their health with two basic options over which they have control: 1) minimize risk of infection by being vaccinated and by wearing a face mask when appropriate, and 2) minimize risk of transmission upon infection by self-isolating. For the latter to be effective, it is essential to have an accurate sense of the probability of infectivity as a function of time following the onset of symptoms. Epidemiological considerations suggest that the period of infectivity follows a lognormal distribution. This proposition is tested empirically by construction of the lognormal probability density function and cumulative distribution function based on quantiles of infectivity reported by several independent investigations. A comprehensive examination of a prototypical ideal clinical study, based on general statistical principles (the Principle of Maximum Entropy and the Central Limit Theorem) reveals that the probability of infectivity is a lognormal random variable. Subsequent evolution of new variants may change the parameters of the distribution, which can be updated by the methods in this paper, but the form of the probability function is expected to remain lognormal as this is the most probable distribution consistent with mathematical requirements and available information.
文摘大规模人脸聚类不仅要求高效的人脸特征,还要求聚类算法在保持高准确率的同时耗时短.本文通过构建卷积神经网络高效提取人脸特征,并采用经典K-means算法和现阶段新颖的CFSFDP(Clustering by Fast Search and Find of Density Peaks)算法进行大规模人脸聚类.实验在聚类数目递增的情况下进行,并通过随机指标(Rand Index,RI)、信息熵、F1-measure和混淆矩阵可视化来综合评估聚类的质量.结果表明,在大规模人脸聚类的情况下,卷积神经网络特征融合K-means的人脸聚类算法速度和准确率均优于CFSFDP算法.这一结论对大规模人脸聚类的实际应用具有重要的指导意义.
