arXiv：cs.LG· Xinhua Miao, Bowei Yang, Zhengong Cai·· 4 小时前
AdaptLSTM：面向分布漂移下云负载预测的高效自适应在线学习
AdaptLSTM: Efficient Adaptive Online Learning for Cloud Workload Forecasting under Distribution Drift
AI 导读
研究者提出 AdaptLSTM，一个通过验证校准阈值检测漂移并做选择性定向更新的自适应在线框架。在 Alibaba Machine Trace 上，它以 20% 成本恢复 Naive Online 54% 的改进（2.7 倍效率）；在波动更大的 Container Trace 上达到 96%（4.8 倍效率，MAE 较 Static 降低 75%）。
正文
Abstract:Accurate workload forecasting is critical for elastic resource provisioning in web-scale cloud services, where distribution shifts driven by viral content, product launches, and user behavior degrade offline-trained models rapidly. Naive online learning recovers accuracy but incurs prohibitive per-step compute cost. We propose AdaptLSTM, an adaptive online framework that detects drift via validation-calibrated thresholds and applies selective, targeted updates. On the Alibaba Machine Trace, AdaptLSTM recovers 54\% of Naive Online's improvement at 20\% cost ($2.7\times$ efficiency, $p=0.002$ over 10 seeds). On the more volatile Container Trace, it achieves 96\% at 20\% cost ($4.8\times$ efficiency, $+75\%$ MAE reduction over Static). Unlike classical drift detectors (ADWIN, DDM, Page-Hinkley) which fail to trigger on regression-scale error streams, AdaptLSTM fires 42 times over 301 steps and outperforms matched-budget baselines. Wall-clock profiling shows $1.33\times$ throughput gain and 45\% update-time reduction. The framework is model-agnostic: identical Pareto patterns hold for LSTM, GRU, and Transformer backbones.
| Subjects: | Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.12265 [cs.LG] |
| (or arXiv:2610.12265v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.12265 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Xinhua Miao [view email]
[v1]
Thu, 8 Oct 2026 16:31:56 UTC (707 KB)
来源：arXiv：cs.LG · arxiv.org