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 2510.03358cs.LGcs.AI
收藏

理解时间序列领域的Transformer:秩结构、秩流与可压缩性

Understanding Transformers for Time Series: Rank Structure, Flow-of-ranks, and Compressibility

Center for Applied Mathematics(应用数学中心);Cornell University(康奈尔大学);Amazon Web Services(亚马逊网络服务);Amazon Selling Partner Services(亚马逊销售合作伙伴服务);Amazon Supply Chain Optimization Technologies(亚马逊供应链优化技术)

Annan Yu, Danielle C. Maddix, Boran Han, Xiyuan Zhang, Abdul Fatir Ansari, Oleksandr Shchur, Christos Faloutsos, Andrew Gordon Wilson, Michael W. Mahoney, Yuyang Wang

AI总结:

该研究从秩结构视角分析时间序列Transformer,提出秩流概念解释秩随深度增长的规律,据此压缩时间序列基础模型Chronos,在无精度损失下大幅降低推理时间与内存占用,为模型设计提供指导。

摘要:

Transformer已被广泛应用于多种数据模态,但从文本模型中提炼出的原理往往无法完美迁移到其他模态的训练模型上。本文从秩结构的视角分析Transformer,聚焦于数据结构特性与文本、视觉领域差异显著的时间序列场景。研究表明,与文本或视觉嵌入不同,时间序列嵌入的奇异值谱呈现急剧衰减的特征:较小的补丁(patch)尺寸与平滑的连续映射会将数据集中到低秩子空间中。基于此,我们证明了对应的$Q/K/V$投影可实现精准的低秩近似,且注意力层的可压缩程度与嵌入谱的衰减程度成正比。我们提出了“秩流(flow-of-ranks)”概念,即跨深度的非线性混合会抬升秩,这解释了为何早期层最适合压缩,以及秩为何随深度增加而增长。在这些理论与实证结果的指导下,我们利用这些洞见对大型时间序列基础模型Chronos进行压缩,实现了推理时间减少65%、内存占用减少81%的效果,且无精度损失。我们的发现为时间序列基础模型的宽度、深度与注意力头分配,以及利用其固有可压缩性提供了原则性指导。

英文摘要:

Transformers are widely used across data modalities, and yet the principles distilled from text models often transfer imperfectly to models trained to other modalities. In this paper, we analyze Transformers through the lens of rank structure. Our focus is on the time series setting, where the structural properties of the data differ remarkably from those of text or vision. We show that time-series embeddings, unlike text or vision, exhibit sharply decaying singular value spectra: small patch sizes and smooth continuous mappings concentrate the data into low-rank subspaces. From this, we prove that the associated $Q/K/V$ projections admit accurate low-rank approximations, and that attention layers become compressible in proportion to the decay of the embedding spectrum. We introduce the concept of flow-of-ranks, a phenomenon by which nonlinear mixing across depth inflates the rank, explaining why early layers are most amenable to compression and why ranks grow with depth. Guided by these theoretical and empirical results, we use these insights to compress Chronos, a large time series foundation model, achieving a reduction of $65\%$ in inference time and $81\%$ in memory, without loss of accuracy. Our findings provide principled guidance for allocating width, depth, and heads in time series foundation models, and for exploiting their inherent compressibility.

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。

↑