Keywords: $f-$divergence, robustness, learning with noisy labels
Abstract: We show when maximizing a properly defined $f$-divergence measure with respect to a classifier's predictions and the supervised labels is robust with label noise. Leveraging its variational form, we derive a nice decoupling property for a family of $f$-divergence measures when label noise presents, where the divergence is shown to be a linear combination of the variational difference defined on the clean distribution and a bias term introduced due to the noise. The above derivation helps us analyze the robustness of different $f$-divergence functions. With established robustness, this family of $f$-divergence functions arises as useful metrics for the problem of learning with noisy labels, which do not require the specification of the labels' noise rate. When they are possibly not robust, we propose fixes to make them so. In addition to the analytical results, we present thorough experimental evidence. Our code is available at https://github.com/UCSC-REAL/Robust-f-divergence-measures.
One-sentence Summary: We show when maximizing a properly defined $f$-divergence measure with respect to a classifier's predictions and the supervised labels is robust with label noise.
Supplementary Material: zip
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics
Code: [![github](/images/github_icon.svg) weijiaheng/Robust-f-divergence-measures](https://github.com/weijiaheng/Robust-f-divergence-measures)
Data: [CIFAR-10](https://paperswithcode.com/dataset/cifar-10), [CIFAR-100](https://paperswithcode.com/dataset/cifar-100), [CIFAR-100N](https://paperswithcode.com/dataset/cifar-100n), [CIFAR-10N](https://paperswithcode.com/dataset/cifar-10n), [Clothing1M](https://paperswithcode.com/dataset/clothing1m), [Fashion-MNIST](https://paperswithcode.com/dataset/fashion-mnist), [MNIST](https://paperswithcode.com/dataset/mnist)