Efficient CNN Compression via Multi-method Low Rank Factorization and Feature Map Similarity

Abstract--Low-Rank Factorization (LRF) is a widely adopted technique for compressing deep neural networks (DNNs). However, it faces several challenges, including optimal rank selection, a vast design space, long fine-tuning times, and limited compatibility with different layer types and decomposition methods. This paper presents an end-to-end Design Space Exploration (DSE) methodology and framework for compressing convolu-tional neural networks (CNNs) that addresses all these issues. We introduce a novel rank selection strategy based on feature map similarity, which captures non-linear interactions between layer outputs more effectively than traditional weight-based approaches. Unlike prior works, our method uses a one-shot fine-tuning process, significantly reducing the overall fine-tuning time. The proposed framework is fully compatible with all types of convolutional (Conv) and fully connected (FC) layers. T o further improve compression, the framework integrates three different LRF techniques for Conv layers and three for FC layers, applying them selectively on a per-layer basis. We demonstrate that combining multiple LRF methods within a single model yields better compression results than using a single method uniformly across all layers. Finally, we provide a comprehensive evaluation and comparison of the six LRF techniques, offering practical insights into their effectiveness across different scenarios. The proposed work is integrated into T ensorFlow 2.x, ensuring compatibility with widely used deep learning workflows. Experimental results on 14 CNN models across eight datasets demonstrate that the proposed methodology achieves substantial compression with minimal accuracy loss, outperforming several state-of-the-art techniques. Impact Statement--Deep neural network (DNN) compression is a critical challenge in artificial intelligence (AI), especially for deploying models in resource-constrained environments. This article addresses the limitations of existing low-rank factorization (LRF) methods, such as suboptimal rank selection, long fine-tuning times, and limited applicability across different layers. A novel end-to-end compression framework is proposed, featuring a feature map similarity-based rank selection strategy, one-shot fine-tuning, and hybrid decomposition support. Unlike prior works, the framework applies different LRF methods per layer and supports six decomposition algorithms across convolutional and fully connected layers. Integrated into T ensorFlow 2.x, it achieves superior compression and accuracy trade-offs across 14 CNN models and eight datasets, outperforming state-of-the-art techniques such as V ariational Bayesian Matrix Factorization and filter-based pruning. This contribution enables scalable, architecture-agnostic compression and offers a practical tool for accelerating DNN deployment in mobile AI, embedded systems, and edge computing scenarios.