Convolution Filter Compression via Sparse Linear Combinations of Quantized Basis
Open Access
- 24 September 2024
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Networks and Learning Systems
- Vol. 36 (6) , 10257-10270
Abstract
Convolutional neural networks (CNNs) have achieved significant performance on various real-life tasks. However, the large number of parameters in convolutional layers requires huge storage and computation resources, making it challenging to deploy CNNs on memory-constrained embedded devices. In this article, we propose a novel compression method that generates the convolution filters in each layer using a set of learnable low-dimensional quantized filter bases. The proposed method reconstructs the convolution filters by stacking the linear combinations of these filter bases. By using quantized values in weights, the compact filters can be represented using fewer bits so that the network can be highly compressed. Furthermore, we explore the sparsity of coefficients through $L_1$ -ball projection when conducting linear combination to further reduce the storage consumption and prevent overfitting. We also provide a detailed analysis of the compression performance of the proposed method. Evaluations of image classification and object detection tasks using various network structures demonstrate that the proposed method achieves a higher compression ratio with comparable accuracy compared with the existing state-of-the-art filter decomposition and network quantization methods.Keywords
Funding Information
- NSFC/Research Grants Council (RGC) Joint Research Scheme (N_HKBU214/21)
- General Research Fund of RGC (12201321, 12202622, 12201323)
- RGC Senior Research Fellow Scheme (SRFS2324-2S02)
- National Natural Science Foundation of China (61906161)
This publication has 58 references indexed in Scilit:
- AdaBin: Improving Binary Neural Networks with Adaptive Binary Sets2022
- Distribution-Sensitive Information Retention for Accurate Binary Neural NetworkInternational Journal of Computer Vision, 2022
- Training Multi-Bit Quantized and Binarized Networks with a Learnable Symmetric QuantizerIEEE Access, 2021
- Sparsity through evolutionary pruning prevents neuronal networks from overfittingNeural Networks, 2020
- Shift-Net: Image Inpainting via Deep Feature Rearrangement2018
- Bi-Real Net: Enhancing the Performance of 1-Bit CNNs with Improved Representational Capability and Advanced Training Algorithm2018
- LQ-Nets: Learned Quantization for Highly Accurate and Compact Deep Neural Networks2018
- Modulated Convolutional Networks2018
- XNOR-Net: ImageNet Classification Using Binary Convolutional Neural Networks2016
- Deep Residual Learning for Image Recognition2016