arXiv cs.AI· Jesper B. Christensen, Ciaran Bench, Spencer A. Thomas, H\"usn\"u Aslan, David Balslev-Harder, Nadia A. S. Smith, Alessandra Manzin·· 13 小时前AI 评分24
Ctrl-A：基于控制理论的自适应在线数据增强方法
Ctrl-A: Control-Driven Machine Learning using Adaptive Online Data Augmentation
AI 导读
研究者提出 ControlAugment（Ctrl-A），一种借鉴控制理论、在训练中在线调整各增强操作强度分布的图像数据增强算法，无需手动初始化单个增强强度。它通过控制回路架构和"相对操作响应曲线"按操作动态更新，有望减少对模型性能有负面影响的增强操作。
正文 · 原文
Abstract:We introduce ControlAugment (Ctrl-A), an automated data augmentation algorithm for image-vision tasks, which incorporates principles from control theory for online adjustment of augmentation strength distributions during model training. Ctrl-A eliminates the need for initialization of individual augmentation strengths. Instead, augmentation strength distributions are dynamically, and individually, adapted during training based on a control-loop architecture and what we define as relative operation response curves. Using an operation-dependent update procedure provides Ctrl-A with the potential to reduce the use of augmentation operations that negatively impact model performance, alleviating the need to manually engineer augmentation policies for new image classification tasks. Experiments on the CIFAR-10, CIFAR-100, SVHN-core, and Tiny ImageNet benchmark datasets using WideResNet-28-10 and ResNet architectures demonstrate that Ctrl-A provides competitive performance across a range of image classification benchmarks.
| Comments: | 19 pages (13 pages main manuscript), 12 figures (9 in main manuscript). v2 note: Title slightly changed to make it less generic/more descriptive. Experimental section overhauled to demonstrate and exemplify closed-loop control mechanism, experiment with Tiny ImageNet added |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Systems and Control (eess.SY) |
| Cite as: | arXiv:2603.21819 [cs.CV] |
| (or arXiv:2603.21819v2 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2603.21819 arXiv-issued DOI via DataCite |
Submission history
From: Jesper Christensen Dr [view email]
[v1]
Mon, 23 Mar 2026 11:03:02 UTC (1,533 KB)
[v2]
Mon, 26 Oct 2026 19:00:01 UTC (2,043 KB)
来源：arXiv cs.AI · arxiv.org