Improvement of semi-supervised learning in real application scenarios

摘要

Due to the high demand of deep learning for data quantity, semi-supervised learning (SSL) has a very importantapplication prospect because of its successful use of unlabeled data. Existing SSL algorithms have achieved highaccuracy on MINIST, CIFAR-10 and SHVN datasets, and even outperform fully supervised algorithms. However,because the above three datasets have the characteristics of balanced data categories and simple identification taskswhich can’t be ignored for classification problems, the SSL algorithm has uncertainties of effectiveness in the case ofunbalanced datasets and specific recognition tasks. We analyze the datasets and find that the number of “disgust” inexpressions dataset is less than other categories, and so is “discussion” in the classroom action recognition dataset.Therefore, we use a novel SSL model: Deep Co-Training (DCT) model to experiment on the expression recognitiondatabase (FER2013), as well as our own classroom student action database (BNU-LCSAD) and analyze the effectivenessof the algorithm in specific application scenarios. Moreover, we use a training strategy of TSA when train our model tosolve the problem of being easily overfitting which is more likely to occur when data categories are not balanced. Theexperimental results prove the effectiveness of the SSL algorithm in practical application and the significance of usingTSA.