[半监督学习] Temporal Ensembling for Semi-Supervised Learning
论文地址:TemporalEnsemblingforSemi-SupervisedLearning代码地址:https://github.com/s-laine/tempens会议:ICLR2017任务:分类self-ensemblingself-ensembling即自集成,自己集成自己.对于神经网络来说,一个样本如果多次送入网络,能够产生多个模型预测结果,这些结果可以进行融合,同时在不同的ba