4 papers
Useful nonrobust features are ubiquitous in biomedical images
Coenraad Mouton, Randle Rabe, Niklas C. Koser +4
We study whether deep networks for medical imaging learn useful nonrobust features - predictive input patterns that are not human interpretable and highly susceptible to small adve…
Does simple trump complex? Comparing strategies for adversarial robustness in DNNs
William Brooks, Marelie H. Davel, Coenraad Mouton
Deep Neural Networks (DNNs) have shown substantial success in various applications but remain vulnerable to adversarial attacks. This study aims to identify and isolate the compone…
Impact of Batch Normalization on Convolutional Network Representations
Hermanus L. Potgieter, Coenraad Mouton, Marelie H. Davel
Batch normalization (BatchNorm) is a popular layer normalization technique used when training deep neural networks. It has been shown to enhance the training speed and accuracy of…
Is network fragmentation a useful complexity measure?
Coenraad Mouton, Randle Rabe, Daniël G. Haasbroek +3
It has been observed that the input space of deep neural network classifiers can exhibit `fragmentation', where the model function rapidly changes class as the input space is trave…