Learning Pruned Structure and Weights Simultaneously from Scratch: an Attention based Approach

TL;DR

This paper introduces ASWL, a novel attention-based method that simultaneously learns sparse network structures and weights from scratch, resulting in efficient pruning with high accuracy across multiple datasets.

Contribution

The paper presents a new unstructured pruning approach using layer-wise attention to learn sparse structures and weights simultaneously from scratch.

Findings

Achieves higher pruning ratios with maintained accuracy.

Outperforms state-of-the-art pruning methods on MNIST, Cifar10, and ImageNet.

Improves network efficiency and reduces storage requirements.

Abstract

As a deep learning model typically contains millions of trainable weights, there has been a growing demand for a more efficient network structure with reduced storage space and improved run-time efficiency. Pruning is one of the most popular network compression techniques. In this paper, we propose a novel unstructured pruning pipeline, Attention-based Simultaneous sparse structure and Weight Learning (ASWL). Unlike traditional channel-wise or weight-wise attention mechanism, ASWL proposed an efficient algorithm to calculate the pruning ratio through layer-wise attention for each layer, and both weights for the dense network and the sparse network are tracked so that the pruned structure is simultaneously learned from randomly initialized weights. Our experiments on MNIST, Cifar10, and ImageNet show that ASWL achieves superior pruning results in terms of accuracy, pruning ratio and operating efficiency when compared with state-of-the-art network pruning methods.