Image Classification on CIFAR-10 (Pruning Efficiency and Accuracy Tradeoff)
0.24Accuracy ChangeDepGraph
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| DepGraphBackbone=ResNet-56, Speed Up Target=~2.1x2023.01 | 0.24 | 93.53 | 93.77 | 2.11 | |
| DepGraphBackbone=ResNet-56, Speed Up Target=~2.5x2023.01 | 0.11 | 93.53 | 93.64 | 2.57 | |
| C-SGDBackbone=ResNet-562023.01 | 0.05 | 93.39 | 93.44 | 2.55 | |
| PolarBackbone=ResNet-562023.01 | 0.03 | 93.8 | 93.83 | 1.88 | |
| ResRepBackbone=ResNet-562023.01 | 0 | 93.71 | 93.71 | 2.12 | |
| GReg-2Backbone=ResNet-562023.01 | 0 | 93.36 | 93.36 | 2.55 | |
| NISPBackbone=ResNet-562023.01 | -0.03 | — | — | 1.76 | |
| DepGraph w/o SLBackbone=ResNet-56, Speed Up Target=~2.1x2023.01 | -0.07 | 93.53 | 93.46 | 2.11 | |
| HRankBackbone=ResNet-562023.01 | -0.09 | 93.26 | 92.17 | 2 | |
| DepGraph w/o SLBackbone=ResNet-56, Speed Up Target=~2.5x2023.01 | -0.17 | 93.53 | 93.36 | 2.51 | |
| GReg-1Backbone=ResNet-562023.01 | -0.18 | 93.36 | 93.18 | 2.55 | |
| SFPBackbone=ResNet-562023.01 | -0.23 | 93.59 | 93.36 | 2.11 | |
| GeometricBackbone=ResNet-562023.01 | -0.33 | 93.59 | 93.26 | 1.7 | |
| GBNBackbone=ResNet-562023.01 | -0.33 | 93.1 | 92.77 | 2.51 | |
| AMCBackbone=ResNet-562023.01 | -0.9 | 92.8 | 91.9 | 2 | |
| AFPBackbone=ResNet-562023.01 | -0.99 | 93.93 | 92.94 | 2.56 | |
| CPBackbone=ResNet-562023.01 | -1 | 92.8 | 91.8 | 2 |