arXiv：cs.LG· Wan Zhang, Qinjie Lin, Chan Lee, Weijian Li, Han Liu, Kai Zhang·· 16 小时前
Hopformer：面向时间序列预测的同质性追求 Transformer
Hopformer: Homogeneity-Pursuit Transformer for Time Series Forecasting
AI 导读
Hopformer 是一个用于高维协变量多时间序列预测的两阶段框架，第一阶段通过 Sparsity Pattern Aggregation（SPA）提取含协变量的低方差共同趋势作为同质化层，第二阶段用 LoRA 微调的 Transformer 建模残差中的复杂依赖。
正文
Abstract:Forecasting multiple time-series with high-dimensional covariates presents a core challenge: unifying common temporal patterns while retaining meaningful series-specific information. We introduce Hopformer (Homogeneity-Pursuit Transformer), a two-stage framework that addresses this challenge. In the first stage, we perform a Sparsity Pattern Aggregation (SPA) scheme extracting a common low-variance trend that incorporates the covariates. This acts as a homogenization layer. In the second stage, a LoRA-fine-tuned Transformer models the remaining complex dependencies in the residual. Our method is theoretically grounded. We prove that SPA achieves a near-optimal bias-variance trade-off via an oracle inequality. We also provide generalization bounds for the second stage under dependent time series data. Hopformer sets a new state of the art, improving MASE by an average of 6.56% across synthetic and real-world forecasting benchmarks.
| Comments: | 36 pages, 5 figures, 25 tables. v2: new end-to-end risk bound under temporal dependence; expanded experiments (6 datasets, 10 baselines) with ablations and statistical inference; extended appendices |
| Subjects: | Machine Learning (stat.ML); Machine Learning (cs.LG) |
| Cite as: | arXiv:2607.22299 [stat.ML] |
| (or arXiv:2607.22299v2 [stat.ML] for this version) | |
| https://doi.org/10.48550/arXiv.2607.22299 arXiv-issued DOI via DataCite |
Submission history
From: Wan Zhang [view email]
[v1]
Fri, 24 Jul 2026 13:44:36 UTC (10,157 KB)
[v2]
Tue, 13 Oct 2026 18:04:05 UTC (4,480 KB)
来源：arXiv：cs.LG · arxiv.org