面部动作单元(Action Unit,AU)识别是计算机视觉与情感计算领域的热点课题.AU识别属于多标签二分类任务,目前面临着标签不均衡等挑战.现有的主流算法利用AU之间的关联,通过调整采样率和AU的权重来进行标签重均衡化.然而,这些方法仅仅使...面部动作单元(Action Unit,AU)识别是计算机视觉与情感计算领域的热点课题.AU识别属于多标签二分类任务,目前面临着标签不均衡等挑战.现有的主流算法利用AU之间的关联,通过调整采样率和AU的权重来进行标签重均衡化.然而,这些方法仅仅使模型预测时从偏向出现频率高的标签转为偏向出现频率低的标签,并未解决偏置问题.根据出现频率的高低可将AU划分为头类和尾类,公平对待每一类是实现AU无偏识别的关键.本文引入因果推理理论,提出基于因果干预的无偏化方法(Causal Intervention for Unbiased facial action unit recognition,CIU),以解决多AU间不均衡的问题.通过调整不平衡域和平衡但不可见域上的经验风险实现模型的无偏性.大量实验结果表明,本方法在基准数据集BP4D、DISFA上超越已有的方法,其中在DISFA上超越当前最先进方法1.1%,且可以学习到无偏的特征表示.展开更多
This study uses <span style="font-family:Verdana;">an empirical</span><span style="font-family:Verdana;"> analysis to quantify the downstream analysis effects of data pre-processi...This study uses <span style="font-family:Verdana;">an empirical</span><span style="font-family:Verdana;"> analysis to quantify the downstream analysis effects of data pre-processing choices. Bootstrap data simulation is used to measure the bias-variance decomposition of an empirical risk function, mean square error (MSE). Results of the risk function decomposition are used to measure the effects of model development choices on </span><span style="font-family:Verdana;">model</span><span style="font-family:Verdana;"> bias, variance, and irreducible error. Measurements of bias and variance are then applied as diagnostic procedures for model pre-processing and development. Best performing model-normalization-data structure combinations were found to illustrate the downstream analysis effects of these model development choices. </span><span style="font-family:Verdana;">In addition</span><span style="font-family:Verdana;">s</span><span style="font-family:Verdana;">, results found from simulations were verified and expanded to include additional data characteristics (imbalanced, sparse) by testing on benchmark datasets available from the UCI Machine Learning Library. Normalization results on benchmark data were consistent with those found using simulations, while also illustrating that more complex and/or non-linear models provide better performance on datasets with additional complexities. Finally, applying the findings from simulation experiments to previously tested applications led to equivalent or improved results with less model development overhead and processing time.</span>展开更多
Some properties of Sugeno measure are further discussed, which is a kind of typical nonadditive measure. The definitions and properties of gλ random variable and its distribution function, expected value, and varianc...Some properties of Sugeno measure are further discussed, which is a kind of typical nonadditive measure. The definitions and properties of gλ random variable and its distribution function, expected value, and variance are then presented. Markov inequality, Chebyshev＇s inequality and the Khinchine＇s Law of Large Numbers on Sugeno measure space are also proven. Furthermore, the concepts of empirical risk functional, expected risk functional and the strict consistency of ERM principle on Sugeno measure space are proposed. According to these properties and concepts, the key theorem of learning theory, the bounds on the rate of convergence of learning process and the relations between these bounds and capacity of the set of functions on Sugeno measure space are given.展开更多
文摘面部动作单元(Action Unit,AU)识别是计算机视觉与情感计算领域的热点课题.AU识别属于多标签二分类任务,目前面临着标签不均衡等挑战.现有的主流算法利用AU之间的关联,通过调整采样率和AU的权重来进行标签重均衡化.然而,这些方法仅仅使模型预测时从偏向出现频率高的标签转为偏向出现频率低的标签,并未解决偏置问题.根据出现频率的高低可将AU划分为头类和尾类,公平对待每一类是实现AU无偏识别的关键.本文引入因果推理理论,提出基于因果干预的无偏化方法(Causal Intervention for Unbiased facial action unit recognition,CIU),以解决多AU间不均衡的问题.通过调整不平衡域和平衡但不可见域上的经验风险实现模型的无偏性.大量实验结果表明,本方法在基准数据集BP4D、DISFA上超越已有的方法,其中在DISFA上超越当前最先进方法1.1%,且可以学习到无偏的特征表示.
文摘This study uses <span style="font-family:Verdana;">an empirical</span><span style="font-family:Verdana;"> analysis to quantify the downstream analysis effects of data pre-processing choices. Bootstrap data simulation is used to measure the bias-variance decomposition of an empirical risk function, mean square error (MSE). Results of the risk function decomposition are used to measure the effects of model development choices on </span><span style="font-family:Verdana;">model</span><span style="font-family:Verdana;"> bias, variance, and irreducible error. Measurements of bias and variance are then applied as diagnostic procedures for model pre-processing and development. Best performing model-normalization-data structure combinations were found to illustrate the downstream analysis effects of these model development choices. </span><span style="font-family:Verdana;">In addition</span><span style="font-family:Verdana;">s</span><span style="font-family:Verdana;">, results found from simulations were verified and expanded to include additional data characteristics (imbalanced, sparse) by testing on benchmark datasets available from the UCI Machine Learning Library. Normalization results on benchmark data were consistent with those found using simulations, while also illustrating that more complex and/or non-linear models provide better performance on datasets with additional complexities. Finally, applying the findings from simulation experiments to previously tested applications led to equivalent or improved results with less model development overhead and processing time.</span>
基金supported by the National Natural Science Foundation of China(Grant No.60573069)the Natural Science Foundation of Hebei Province(Grant No.F2004000129)+1 种基金the Key Scientific Research Project of Hebei Education Department(Grant No.2005001D)the Key Scientific and Technical Research Project of the Ministry of Education of China(Grant No.20602).
文摘Some properties of Sugeno measure are further discussed, which is a kind of typical nonadditive measure. The definitions and properties of gλ random variable and its distribution function, expected value, and variance are then presented. Markov inequality, Chebyshev＇s inequality and the Khinchine＇s Law of Large Numbers on Sugeno measure space are also proven. Furthermore, the concepts of empirical risk functional, expected risk functional and the strict consistency of ERM principle on Sugeno measure space are proposed. According to these properties and concepts, the key theorem of learning theory, the bounds on the rate of convergence of learning process and the relations between these bounds and capacity of the set of functions on Sugeno measure space are given.
