SOTAVerified

Network Pruning

Network Pruning is a popular approach to reduce a heavy network to obtain a light-weight form by removing redundancy in the heavy network. In this approach, a complex over-parameterized network is first trained, then pruned based on come criterions, and finally fine-tuned to achieve comparable performance with reduced parameters.

Source: Ensemble Knowledge Distillation for Learning Improved and Efficient Networks

Papers

Showing 76100 of 534 papers

TitleStatusHype
A Gradient Flow Framework For Analyzing Network PruningCode1
Sanity-Checking Pruning Methods: Random Tickets can Win the JackpotCode1
Accelerating Sparse DNN Models without Hardware-Support via Tile-Wise SparsityCode1
EagleEye: Fast Sub-net Evaluation for Efficient Neural Network PruningCode1
Paying more attention to snapshots of Iterative Pruning: Improving Model Compression via Ensemble DistillationCode1
Weight Pruning via Adaptive Sparsity LossCode1
Shapley Value as Principled Metric for Structured Network PruningCode1
MicroNet for Efficient Language ModelingCode1
Movement Pruning: Adaptive Sparsity by Fine-TuningCode1
Dynamic Sparse Training: Find Efficient Sparse Network From Scratch With Trainable Masked LayersCode1
Network Adjustment: Channel Search Guided by FLOPs Utilization RatioCode1
DHP: Differentiable Meta Pruning via HyperNetworksCode1
How Not to Give a FLOP: Combining Regularization and Pruning for Efficient InferenceCode1
Similarity of Neural Networks with GradientsCode1
What is the State of Neural Network Pruning?Code1
Comparing Rewinding and Fine-tuning in Neural Network PruningCode1
Good Subnetworks Provably Exist: Pruning via Greedy Forward SelectionCode1
HYDRA: Pruning Adversarially Robust Neural NetworksCode1
HRank: Filter Pruning using High-Rank Feature MapCode1
Knapsack Pruning with Inner DistillationCode1
Picking Winning Tickets Before Training by Preserving Gradient FlowCode1
Soft Threshold Weight Reparameterization for Learnable SparsityCode1
Filter Sketch for Network PruningCode1
Quantisation and Pruning for Neural Network Compression and RegularisationCode1
Sparse Weight Activation TrainingCode1
Show:102550
← PrevPage 4 of 22Next →

Benchmark Results

#ModelMetricClaimedVerifiedStatus
1ResNet50-2.3 GFLOPsAccuracy78.79Unverified
2ResNet50-1.5 GFLOPsAccuracy78.07Unverified
3ResNet50 2.5 GFLOPSAccuracy78Unverified
4RegX-1.6GAccuracy77.97Unverified
5ResNet50 2.0 GFLOPSAccuracy77.7Unverified
6ResNet50-3G FLOPsAccuracy77.1Unverified
7ResNet50-2G FLOPsAccuracy76.4Unverified
8ResNet50-1G FLOPsAccuracy76.38Unverified
9TAS-pruned ResNet-50Accuracy76.2Unverified
10ResNet50Accuracy75.59Unverified
#ModelMetricClaimedVerifiedStatus
1FeatherTop-1 Accuracy76.93Unverified
2SpartanTop-1 Accuracy76.17Unverified
3ST-3Top-1 Accuracy76.03Unverified
4AC/DCTop-1 Accuracy75.64Unverified
5CSTop-1 Accuracy75.5Unverified
6ProbMaskTop-1 Accuracy74.68Unverified
7STRTop-1 Accuracy74.31Unverified
8DNWTop-1 Accuracy74Unverified
9GMPTop-1 Accuracy73.91Unverified
#ModelMetricClaimedVerifiedStatus
1+U-DML*Inference Time (ms)675.56Unverified
2DenseAccuracy79Unverified
3AC/DCAccuracy78.2Unverified
4Beta-RankAccuracy74.01Unverified
5TAS-pruned ResNet-110Accuracy73.16Unverified
#ModelMetricClaimedVerifiedStatus
1TAS-pruned ResNet-110Accuracy94.33Unverified
2ShuffleNet – QuantisedInference Time (ms)23.15Unverified
3AlexNet – QuantisedInference Time (ms)5.23Unverified
4MobileNet – QuantisedInference Time (ms)4.74Unverified
#ModelMetricClaimedVerifiedStatus
1FFN-ShapleyPrunedAvg #Steps12.05Unverified