Robust Deep Graph Based Learning for Binary Classification
This work addresses the challenge of noisy labels in binary classification, which is a common issue in real-world datasets, but it appears to be an incremental improvement over existing robust methods.
The paper tackles the problem of binary classification with noisy training labels by proposing a robust CNN-based classifier that uses deep metric functions to construct an optimal graph for cleaning labels via graph Laplacian regularization. The method outperforms several state-of-the-art classifiers on three datasets, demonstrating improved performance in noisy conditions.
Convolutional neural network (CNN)-based feature learning has become state of the art, since given sufficient training data, CNN can significantly outperform traditional methods for various classification tasks. However, feature learning becomes more difficult if some training labels are noisy. With traditional regularization techniques, CNN often overfits to the noisy training labels, resulting in sub-par classification performance. In this paper, we propose a robust binary classifier, based on CNNs, to learn deep metric functions, which are then used to construct an optimal underlying graph structure used to clean noisy labels via graph Laplacian regularization (GLR). GLR is posed as a convex maximum a posteriori (MAP) problem solved via convex quadratic programming (QP). To penalize samples around the decision boundary, we propose two regularized loss functions for semi-supervised learning. The binary classification experiments on three datasets, varying in number and type of features, demonstrate that given a noisy training dataset, our proposed networks outperform several state-of-the-art classifiers, including label-noise robust support vector machine, CNNs with three different robust loss functions, model-based GLR, and dynamic graph CNN classifiers.