Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Soft Threshold Weight Reparameterization for Learnable Sparsity

About

Sparsity in Deep Neural Networks (DNNs) is studied extensively with the focus of maximizing prediction accuracy given an overall parameter budget. Existing methods rely on uniform or heuristic non-uniform sparsity budgets which have sub-optimal layer-wise parameter allocation resulting in a) lower prediction accuracy or b) higher inference cost (FLOPs). This work proposes Soft Threshold Reparameterization (STR), a novel use of the soft-threshold operator on DNN weights. STR smoothly induces sparsity while learning pruning thresholds thereby obtaining a non-uniform sparsity budget. Our method achieves state-of-the-art accuracy for unstructured sparsity in CNNs (ResNet50 and MobileNetV1 on ImageNet-1K), and, additionally, learns non-uniform budgets that empirically reduce the FLOPs by up to 50%. Notably, STR boosts the accuracy over existing results by up to 10% in the ultra sparse (99%) regime and can also be used to induce low-rank (structured sparsity) in RNNs. In short, STR is a simple mechanism which learns effective sparsity budgets that contrast with popular heuristics. Code, pretrained models and sparsity budgets are at https://github.com/RAIVNLab/STR.

Aditya Kusupati, Vivek Ramanujan, Raghav Somani, Mitchell Wortsman, Prateek Jain, Sham Kakade, Ali Farhadi• 2020

Related benchmarks

TaskDatasetResultRank
Image ClassificationCIFAR-100 (test)
Accuracy73.45
3518
Image ClassificationCIFAR-10 (test)
Accuracy93.73
3381
Image ClassificationImageNet-1K 1.0 (val)
Top-1 Accuracy76.19
1866
Image ClassificationImageNet-1k (val)
Top-1 Accuracy76.19
1453
Image ClassificationImageNet (val)
Top-1 Acc76.19
1206
Image ClassificationImageNet-1k (val)
Top-1 Accuracy62.1
840
Image ClassificationImageNet ILSVRC-2012 (val)
Top-1 Accuracy76.19
405
Image ClassificationImageNet (val)
Accuracy76
300
Image ClassificationImageNet (val)
Top-1 Accuracy62.1
118
Natural Language InferenceMNLI (matched)
Accuracy75.8
110
Showing 10 of 16 rows

Other info

Code

Follow for update