%0 Journal Article
%T 基于多标签随机游走的选择性集成方法用于表情识别<br>A Selective Ensemble Method Based on Multi-Label Random Walk for Expression Recognition
%A 黄仕松
%A 李丹杨
%A 陈星
%A 唐玉梅
%A 吴义青
%J Software Engineering and Applications
%P 1344-1356
%@ 2325-2278
%D 2022
%I Hans Publishing
%R 10.12677/SEA.2022.116138
%X 为了提升分类器集成的性能，本文提出了一种基于多标签随机游走的选择性集成方法，该方法将分类器选择问题建模为多标签分类问题，以灵活有效选择分类器。首先在训练集样本与分类器序列间建立映射，将能正确识别样本的分类器序列视为样本标签，对于一个特定的测试样本，寻找它的近邻样本，并构建出多标签随机游走图，执行随机游走过程，根据收敛后的概率向量选择出对应的分类器序列进行集成。在Fer2013、CK+和JAFFE人脸表情数据集上进行实验，并与一些当前先进的选择性集成算法进行对比，实验结果证明了该算法的可行性和有效性。<br />
In order to improve the ensemble performance of classifier ensemble, a selective ensemble method based on multi-label random walk is proposed, which models the classifier selection problem as a multi-label classification problem to select classifiers flexibly and effectively. First, a mapping is established between the training set samples and the classifier sequence, and the classifier sequence that can correctly identify the sample is regarded as the sample label. For a specific test sample, its neighbor samples are found, and a multi-label random walk graph is constructed. A random walk process is performed, and the corresponding classifier sequence is selected for ensemble according to the converged probability vector. Experiments are carried out on Fer2013, CK+ and JAFFE facial expression datasets, and compared with some current advanced selective ensemble algorithms. The experimental results demonstrate the feasibility and effectiveness of the algorithm.
%K 选择性集成，随机游走，卷积神经网络，人脸表情识别<br>Selective Ensemble
%K Random Walk
%K Convolutional Neural Network
%K Expression Recognition
%U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=59657