arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2607.08754cs.LGcs.AI
收藏

SLORR:简单高效的训练中低秩正则化

SLORR: Simple and Efficient In-Training Low-Rank Regularization

Max Planck Institute for Intelligent Systems(马克斯·普朗克智能系统研究所);University of Tübingen(图宾根大学);ELLIS Institute Tübingen(图宾根ELLIS研究所);Tübingen AI Center(图宾根人工智能中心)

David González-Martínez, Shiwei Liu

AI总结:

研究针对神经网络低秩正则化难题,提出简单无状态的SLORR框架,基于霍耶尔稀疏性度量和核范数有两个变体,通过GPU友好近似正则化权重矩阵,在ImageNet-1K及LLM预训练中验证其能在低开销下诱导可压缩性并保持性能。

摘要:

低秩分解广泛用于压缩神经网络,但现代模型往往难以在不显著损失精度的情况下进行激进分解。现有训练时低秩正则化方法存在局限。为此引入SLORR,一种简单、无状态且保持架构的训练中低秩正则化框架,有基于霍耶尔稀疏性度量和核范数的两个主要变体。通过GPU友好近似直接对原始权重矩阵正则化,给出近似保证。在ImageNet-1K和LLM预训练中评估,结果显示其在引入较少训练开销时能诱导可压缩性并保持性能。

英文摘要:

Low-rank factorization is widely used to compress neural networks, but modern models are often not naturally amenable to aggressive factorization without significant accuracy loss. Existing training-time low-rank regularizers can improve compressibility, but they often require SVDs of large weight matrices, modify the model architecture (introducing additional trainable parameters), or rely on stateful cached quantities. To address these limitations, we introduce SLORR, a simple, stateless, and architecture-preserving framework for in-training low-rank regularization, instantiated with two main variants based on the Hoyer sparsity metric and the nuclear norm. SLORR directly regularizes the original weight matrices using GPU-friendly approximations for the forward and backward passes of the regularizers, for which we provide approximation guarantees. We first evaluate SLORR on ImageNet-1K across short-horizon continued training of ResNet-50, ViT-B/16, and ViT-L/16, and pretraining of ResNet-18, where SLORR induces compressibility while introducing less than 8% training overhead. We further evaluate SLORR-Hoyer in LLM pretraining at 135M and 560M scales: SLORR-trained compressed models preserve performance substantially better than unregularized models while adding less than 1% average training overhead.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑