VNE: An Effective Method for Improving Deep Representation by Manipulating Eigenvalue Distribution

Open paper
Benchmark Model Rank Results
domain-generalization-on-office-homeVNE (ResNet-50, SWAD)#24Average Accuracy: 71.1
domain-generalization-on-pacs-2VNE (ResNet-50, SWAD)#27Average Accuracy: 88.3
domain-generalization-on-terraincognitaVNE (ResNet-50, SWAD)#16Average Accuracy: 51.7
domain-generalization-on-vlcsVNE (ResNet-50, SWAD)#18Average Accuracy: 79.7
few-shot-image-classification-on-mini-2VNE (BOIL)#88Accuracy: 50.95
few-shot-image-classification-on-mini-3VNE (BOIL)#80Accuracy: 67.52
self-supervised-image-classification-on-imagenetI-VNE+ (ResNet-50)#88Top 1 Accuracy: 72.1Top 5 Accuracy: 91.0Number of Params: 25M
semi-supervised-image-classification-on-1I-VNE+ (ResNet-50)#40Top 1 Accuracy: 55.8Top 5 Accuracy: 81.0
semi-supervised-image-classification-on-2I-VNE+ (ResNet-50)#37Top 1 Accuracy: 69.1Top 5 Accuracy: 89.9

Rank counts only results with a code link.