Channel Pruning for Accelerating Very Deep Neural Networks

TL;DR

This paper presents a novel channel pruning technique for deep CNNs that significantly accelerates models like VGG-16, ResNet, and Xception with minimal accuracy loss, enabling faster inference.

Contribution

The authors introduce an iterative two-step channel pruning algorithm based on LASSO regression, improving speed while maintaining accuracy across various architectures.

Findings

Achieved 5x speed-up on VGG-16 with only 0.3% accuracy increase.

Accelerated ResNet and Xception with around 1% accuracy loss at 2x speed-up.

Generalized the pruning method to multi-layer and multi-branch networks.

Abstract

In this paper, we introduce a new channel pruning method to accelerate very deep convolutional neural networks.Given a trained CNN model, we propose an iterative two-step algorithm to effectively prune each layer, by a LASSO regression based channel selection and least square reconstruction. We further generalize this algorithm to multi-layer and multi-branch cases. Our method reduces the accumulated error and enhance the compatibility with various architectures. Our pruned VGG-16 achieves the state-of-the-art results by 5x speed-up along with only 0.3% increase of error. More importantly, our method is able to accelerate modern networks like ResNet, Xception and suffers only 1.4%, 1.0% accuracy loss under 2x speed-up respectively, which is significant. Code has been made publicly available.