Self-Adaptive Network Pruning

TL;DR

This paper introduces SANP, a self-adaptive pruning method for CNNs that dynamically adjusts channel pruning per layer and sample to meet computation budgets, improving efficiency and robustness.

Contribution

The paper presents a novel Saliency-and-Pruning Module enabling adaptive, sample-specific pruning in CNNs, outperforming existing methods in accuracy and pruning rate.

Findings

SANP achieves higher accuracy than state-of-the-art methods.

SANP effectively meets computation budgets across datasets.

SANP demonstrates robustness to different datasets and backbone architectures.

Abstract

Deep convolutional neural networks have been proved successful on a wide range of tasks, yet they are still hindered by their large computation cost in many industrial scenarios. In this paper, we propose to reduce such cost for CNNs through a self-adaptive network pruning method (SANP). Our method introduces a general Saliency-and-Pruning Module (SPM) for each convolutional layer, which learns to predict saliency scores and applies pruning for each channel. Given a total computation budget, SANP adaptively determines the pruning strategy with respect to each layer and each sample, such that the average computation cost meets the budget. This design allows SANP to be more efficient in computation, as well as more robust to datasets and backbones. Extensive experiments on 2 datasets and 3 backbones show that SANP surpasses state-of-the-art methods in both classification accuracy and pruning rate.