Network Pruning on ImageNet - ResNet 50 - 90% sparsity
Metric: Top-1 Accuracy (higher is better)
Metric: Top-1 Accuracy (higher is better)
| # | Model↕ | Top-1 Accuracy▼ | Augmentations | Paper | Date↕ | Code |
|---|---|---|---|---|---|---|
| 1 | Feather | 76.93 | No | Feather: An Elegant Solution to Effective DNN Sp... | 2023-10-03 | Code |
| 2 | Spartan | 76.17 | No | Spartan: Differentiable Sparsity via Regularized... | 2022-05-27 | Code |
| 3 | ST-3 | 76.03 | No | Are Straight-Through gradients and Soft-Threshol... | 2022-12-02 | Code |
| 4 | AC/DC | 75.64 | No | AC/DC: Alternating Compressed/DeCompressed Train... | 2021-06-23 | Code |
| 5 | CS | 75.5 | No | Winning the Lottery with Continuous Sparsification | 2019-12-10 | Code |
| 6 | ProbMask | 74.68 | No | Effective Sparsification of Neural Networks with... | 2021-05-03 | Code |
| 7 | STR | 74.31 | No | Soft Threshold Weight Reparameterization for Lea... | 2020-02-08 | Code |
| 8 | DNW | 74 | No | Discovering Neural Wirings | 2019-06-03 | Code |
| 9 | GMP | 73.91 | No | Soft Threshold Weight Reparameterization for Lea... | 2020-02-08 | Code |