When Optimizing $f$-divergence is Robust with Label Noise

Benchmark Model Rank Results
image-classification-on-clothing1mRobust f-divergence#32Accuracy: 73.09%
learning-with-noisy-labels-on-cifar-100nF-div#18Accuracy (mean): 57.10
learning-with-noisy-labels-on-cifar-10nF-div#14Accuracy (mean): 91.64
learning-with-noisy-labels-on-cifar-10n-1F-div#16Accuracy (mean): 89.70
learning-with-noisy-labels-on-cifar-10n-2F-div#14Accuracy (mean): 89.79
learning-with-noisy-labels-on-cifar-10n-3F-div#14Accuracy (mean): 89.55
learning-with-noisy-labels-on-cifar-10n-worstF-div#18Accuracy (mean): 82.53