Deep unsupervised learning through spatial contrasting.
arXiv: Learning(2017)
摘要
Convolutional networks have marked their place over the last few years as thebest performing model for various visual tasks. They are, however, most suitedfor supervised learning from large amounts of labeled data. Previous attemptshave been made to use unlabeled data to improve model performance by applyingunsupervised techniques. These attempts require different architectures and training methods.In this work we present a novel approach for unsupervised trainingof Convolutional networks that is based on contrasting between spatial regionswithin images. This criterion can be employed within conventional neural net-works and trained using standard techniques such as SGD and back-propagation,thus complementing supervised methods.
更多查看译文
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络