A Unified Framework for Soft Threshold Pruning

TL;DR

This paper introduces a unified theoretical framework for soft threshold pruning using ISTA, leading to an optimal scheduler and state-of-the-art results across various neural network models and datasets.

Contribution

It reformulates soft threshold pruning as an implicit optimization problem, deriving an optimal threshold scheduler and unifying previous methods under a common theoretical perspective.

Findings

Achieved state-of-the-art pruning performance on ImageNet with ResNet-50 and MobileNet-V1.

Unified various pruning strategies under a single ISTA-based framework.

Demonstrated the framework's applicability to different neural network types, including SNNs.

Abstract

Soft threshold pruning is among the cutting-edge pruning methods with state-of-the-art performance. However, previous methods either perform aimless searching on the threshold scheduler or simply set the threshold trainable, lacking theoretical explanation from a unified perspective. In this work, we reformulate soft threshold pruning as an implicit optimization problem solved using the Iterative Shrinkage-Thresholding Algorithm (ISTA), a classic method from the fields of sparse recovery and compressed sensing. Under this theoretical framework, all threshold tuning strategies proposed in previous studies of soft threshold pruning are concluded as different styles of tuning $L_1$-regularization term. We further derive an optimal threshold scheduler through an in-depth study of threshold scheduling based on our framework. This scheduler keeps $L_1$-regularization coefficient stable, implying a time-invariant objective function from the perspective of optimization. In principle, the derived pruning algorithm could sparsify any mathematical model trained via SGD. We conduct extensive experiments and verify its state-of-the-art performance on both Artificial Neural Networks (ResNet-50 and MobileNet-V1) and Spiking Neural Networks (SEW ResNet-18) on ImageNet datasets. On the basis of this framework, we derive a family of pruning methods, including sparsify-during-training, early pruning, and pruning at initialization. The code is available at https://github.com/Yanqi-Chen/LATS.