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

2025-07-30

AI Summary

To address the challenge of deploying deep neural networks (DNNs) on edge devices—where models are often overly parameter-heavy and existing compression techniques struggle to balance accuracy and efficiency—this paper proposes the Fractional Gaussian Filtering and Pruning (FGFP) framework. FGFP introduces a novel fractional Gaussian convolution kernel, uniquely integrating fractional-order calculus with the Gaussian function, containing only seven learnable parameters. Computational complexity is reduced via the Grünwald–Letnikov difference approximation, and fine-grained sparsification is achieved through an adaptive unstructured pruning (AUP) strategy. The method achieves superior accuracy–compression trade-offs while maintaining extreme model lightness. Experiments demonstrate that, on CIFAR-10, FGFP compresses ResNet-20 by 85.2% with only a 1.52% accuracy drop; on ImageNet2012, it reduces ResNet-50 size by 69.1% while incurring merely a 1.63% top-1 accuracy degradation. FGFP consistently outperforms state-of-the-art model compression approaches in overall effectiveness.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Computer Vision: Learning & Optimization for CV
• Natural Language Processing: Learning & Optimization for NLP

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

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 Grünwald-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%.

Problem

• Compress DNNs for edge devices efficiently
• Reduce computational complexity in neural networks
• Achieve high accuracy with minimal model size

Innovation

• Fractional Gaussian filters for DNN compression
• Grünwald-Letnikov derivatives simplify fractional calculus
• Adaptive Unstructured Pruning boosts compression ratios