Averaging Weights Leads to Wider Optima and Better Generalization

Open paper
Benchmark Model Rank Results
image-classification-on-cifar-100PyramidNet-272 + SWA#104Percentage correct: 84.16
image-classification-on-cifar-100WRN+SWA#139Percentage correct: 82.15
image-classification-on-cifar-10ShakeShake-2x64d + SWA#114Percentage correct: 97.12
image-classification-on-cifar-10WRN-28-10 + SWA#124Percentage correct: 96.79
image-classification-on-imagenetResNet-152 + SWA#836Top 1 Accuracy: 78.94%
image-classification-on-imagenetDenseNet-161 + SWA#871Top 1 Accuracy: 78.44%

Rank counts only results with a code link.