LoRA-Edge让边缘设备高效微调CNN，参数量减少98%仍保持高精度。
LoRA-Edge: Tensor-Train-Assisted LoRA for Practical CNN Fine-Tuning on Edge Devices
- 用张量分解压缩卷积层，仅更新输出侧核心参数
- 在多个动作识别数据集上逼近全微调精度，参数更新率≤1.49%
- 适合资源受限的边缘设备，收敛速度提升1.4~3.8倍
在边缘应用如人体行为识别（HAR）中，本地微调卷积神经网络对应对领域偏移至关重要，但全量微调受内存、计算和能耗严格限制。本文提出LoRA-Edge，一种基于低秩适应（LoRA）并引入张量-列车分解（TT-SVD）的参数高效微调方法。该方法（i）对预训练卷积层施加TT-SVD，（ii）仅对输出侧核心进行零初始化的可选更新，确保辅助路径初始状态无效，（iii）将更新融合回稠密卷积核，推理开销不变。此设计保持卷积结构，相比全微调可减少达两个数量级的可训练参数。在多种HAR数据集与CNN骨干网络上，LoRA-Edge达到距全微调精度最高4.7%的水平，且仅更新最多1.49%的参数，持续优于同类参数高效基线。在Jetson Orin Nano上，TT-SVD初始化与选择性核心训练使收敛速度提升1.4至3.8倍。因此，该方法使结构对齐的、参数高效的边缘端CNN适配成为可能。
原文摘要 · Abstract (English)
On-device fine-tuning of CNNs is essential to withstand domain shift in edge applications such as Human Activity Recognition (HAR), yet full fine-tuning is infeasible under strict memory, compute, and energy budgets. We present LoRA-Edge, a parameter-efficient fine-tuning (PEFT) method that builds on Low-Rank Adaptation (LoRA) with tensor-train assistance. LoRA-Edge (i) applies Tensor-Train Singular Value Decomposition (TT-SVD) to pre-trained convolutional layers, (ii) selectively updates only the output-side core with zero-initialization to keep the auxiliary path inactive at the start, and (iii) fuses the update back into dense kernels, leaving inference cost unchanged. This design preserves convolutional structure and reduces the number of trainable parameters by up to two orders of magnitude compared to full fine-tuning. Across diverse HAR datasets and CNN backbones, LoRA-Edge achieves accuracy within 4.7% of full fine-tuning while updating at most 1.49% of parameters, consistently outperforming prior parameter-efficient baselines under similar budgets. On a Jetson Orin Nano, TT-SVD initialization and selective-core training yield 1.4-3.8x faster convergence to target F1. LoRA-Edge thus makes structure-aligned, parameter-efficient on-device CNN adaptation practical for edge platforms.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。