Knowledge Distillation on CIFAR-10 + CIFAR-100
94AccuracyMDA
Evaluation results
| Method | Links | |
|---|---|---|
| MDABackbone=VGG-16, Category=0+02022.03 | 94 | |
| Base (Source models average)Backbone=VGG-16, Category=0+02022.03 | 89.2 | |
| Knowledge Amalgamating (KA)Backbone=VGG-16, Category=0+02022.03 | 84.35 | |
| Base (Source models average)Backbone=VGG-16, Category=3-9 + 20-692022.03 | 75.16 | |
| Base (Source models average)Backbone=VGG-16, Category=0-9 + 20-992022.03 | 74.87 | |
| MDABackbone=VGG-16, Category=0-2 + 0-192022.03 | 74.17 | |
| Base (Source models average)Backbone=VGG-16, Category=0-2 + 0-192022.03 | 74.03 | |
| MDABackbone=VGG-16, Category=3-9 + 20-692022.03 | 73.72 | |
| MDABackbone=VGG-16, Category=0-9 + 20-992022.03 | 72.07 | |
| Knowledge Amalgamating (KA)Backbone=VGG-16, Category=0-9 + 20-992022.03 | 71.6 | |
| Knowledge Amalgamating (KA)Backbone=VGG-16, Category=3-9 + 20-692022.03 | 71.38 | |
| Knowledge Amalgamating (KA)Backbone=VGG-16, Category=0-2 + 0-192022.03 | 69.46 |