Poster
Multi-class classification without multi-class labels
Yen-Chang Hsu · Zhaoyang Lv · Joel Schlosser · Phillip Odom · Zsolt Kira
Great Hall BC #80
Keywords: [ cross-task ] [ weak supervision ] [ problem reduction ] [ classification ] [ learning ] [ neural network ] [ semi-supervised learning ] [ deep learning ] [ unsupervised learning ]
This work presents a new strategy for multi-class classification that requires no class-specific labels, but instead leverages pairwise similarity between examples, which is a weaker form of annotation. The proposed method, meta classification learning, optimizes a binary classifier for pairwise similarity prediction and through this process learns a multi-class classifier as a submodule. We formulate this approach, present a probabilistic graphical model for it, and derive a surprisingly simple loss function that can be used to learn neural network-based models. We then demonstrate that this same framework generalizes to the supervised, unsupervised cross-task, and semi-supervised settings. Our method is evaluated against state of the art in all three learning paradigms and shows a superior or comparable accuracy, providing evidence that learning multi-class classification without multi-class labels is a viable learning option.
Live content is unavailable. Log in and register to view live content