FGFP: A Fractional Gaussian Filter and Pruning for Deep Neural Networks Compression

TL;DR

This paper introduces FGFP, a novel framework combining fractional Gaussian filters and pruning to significantly compress deep neural networks while maintaining high accuracy on benchmarks.

Contribution

The paper proposes a new fractional Gaussian filter design and an adaptive pruning method, achieving high compression ratios with minimal accuracy loss.

Findings

ResNet-20 on CIFAR-10 with 85.2% size reduction and 1.52% accuracy drop.

ResNet-50 on ImageNet with 69.1% size reduction and 1.63% accuracy drop.

Outperforms recent compression methods in accuracy and compression ratio.

Abstract

Network compression techniques have become increasingly important in recent years because the loads of Deep Neural Networks (DNNs) are heavy for edge devices in real-world applications. While many methods compress neural network parameters, deploying these models on edge devices remains challenging. To address this, we propose the fractional Gaussian filter and pruning (FGFP) framework, which integrates fractional-order differential calculus and Gaussian function to construct fractional Gaussian filters (FGFs). To reduce the computational complexity of fractional-order differential operations, we introduce Gr\"unwald-Letnikov fractional derivatives to approximate the fractional-order differential equation. The number of parameters for each kernel in FGF is minimized to only seven. Beyond the architecture of Fractional Gaussian Filters, our FGFP framework also incorporates Adaptive Unstructured Pruning (AUP) to achieve higher compression ratios. Experiments on various architectures and benchmarks show that our FGFP framework outperforms recent methods in accuracy and compression. On CIFAR-10, ResNet-20 achieves only a 1.52% drop in accuracy while reducing the model size by 85.2%. On ImageNet2012, ResNet-50 achieves only a 1.63% drop in accuracy while reducing the model size by 69.1%.