Regularizing Neural Networks via Adversarial Model Perturbation

Open paper
Benchmark Model Rank Results
image-classification-on-svhnPyramidNet + AA (AMP)#7Percentage error: 1.35
image-classification-on-svhnPreActResNet18 (AMP)#27Percentage error: 2.30
image-classification-on-cifar-100PyramidNet + AA (AMP)#75Percentage correct: 86.64
image-classification-on-cifar-100PreActResNet18 (AMP)#168Percentage correct: 78.49
image-classification-on-cifar-10PyramidNet + AA (AMP)#76Percentage correct: 98.02
image-classification-on-cifar-10PreActResNet18 (AMP)#145Percentage correct: 96.03

Rank counts only results with a code link.