Figure 2 - uploaded by Yan Yan
Content may be subject to copyright.
Download
View publication
Prediction histogram of positive (blue) and negative (red) samples for the models trained by AUC-M loss and CE loss on Melanoma training dataset.

Prediction histogram of positive (blue) and negative (red) samples for the models trained by AUC-M loss and CE loss on Melanoma training dataset.

Source publication
Figure 2: Prediction histogram of positive (blue) and negative (red)...
Figure 3: Ablation Study on Batch Score Normalization.
Figure 6: Testing AUC vs epochs on Benchmark Datasets for DenseNet121.
Figure 7: Testing AUC vs epochs on Benchmark Datasets for ResNet20.
+3 Averaged Testing AUC Scores on CheXpert. NBRC means the # of...
Robust Deep AUC Maximization: A New Surrogate Loss and Empirical Studies on Medical Image Classification
Preprint
Full-text available
  • Dec 2020
Deep AUC Maximization (DAM) is a paradigm for learning a deep neural network by maximizing the AUC score of the model on a dataset. Most previous works of AUC maximization focus on the perspective of optimization by designing efficient stochastic algorithms, and studies on generalization performance of DAM on difficult tasks are missing. In this wo...
Cite
Download full-text

Contexts in source publication

Context 1
... addition, the AUC Margin loss is better than AUC Square loss. We also plot the histogram of predictions on training data of our best DAM method (AUC-M+Meta) compared with standard DL method with CE loss in Figure 2. We can see that the predictions made by the DAM method have two well-separated patterns corresponding to positive and negative data. ...
Context 2
... addition, the AUC Margin loss is better than AUC Square loss. We also plot the histogram of predictions on training data of our best DAM method (AUC-M+Meta) compared with standard DL method with CE loss in Figure 2. We can see that the predictions by the DAM method have two well-separated patterns corresponding to positive and negative data. ...