RelativeUncertaintyLearningforFacialExpression RecognitionSupplementaryMaterial AVisualizationresultsonMNISTandCIFAR

Weprovide visualization results onMNIST and CIFAR toshowour uncertainty learning method also works well on datasets besides facial expression recognition (FER) tasks. Weutilize red rectangles to mark images that are misclassified and green rectangles to mark images that are rightly classified. They are usually very hard to be rightlyclassified. We also carry out experiments on MNIST and CIFAR with synthetic noises. If the maximum prediction probability is higher than the one of given label with a threshold (set to 0.2), we believe that sample contains label noise and then change the label to the index of the maximum prediction probability.