<?xml version="1.1" encoding="utf-8"?>
<article xsi:noNamespaceSchemaLocation="http://jats.nlm.nih.gov/publishing/1.1/xsd/JATS-journalpublishing1-mathml3.xsd" dtd-version="1.1" xmlns:xlink="http://www.w3.org/1999/xlink" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"><front><journal-meta><journal-id journal-id-type="publisher-id">TACS</journal-id><journal-title-group><journal-title>Technology and Application of Computer Science</journal-title></journal-title-group><issn>2998-8926</issn><eissn>2998-8934</eissn><publisher><publisher-name>Art and Technology</publisher-name></publisher></journal-meta><article-meta><article-id pub-id-type="doi">10.61369/TACS.2026070002</article-id><article-categories><subj-group subj-group-type="heading"><subject>Article</subject></subj-group></article-categories><title>面向深度学习评估过程的边界自适应插值策略研究</title><url>https://artdesignp.com/journal/TACS/3/7/10.61369/TACS.2026070002</url><author>刘存思,王秋月</author><pub-date pub-type="publication-year"><year>2026</year></pub-date><volume>3</volume><issue>7</issue><history><date date-type="pub"><published-time>2026-04-14</published-time></date></history><abstract>在深度学习模型训练和生成过程中，通常每一轮训练后都会进行模型评估，人脸识别模型训练的早期轮次为例，模型评估过程有概率会因插值边界处理不当而训练程序崩溃，导致训练鲁棒性不够。本文深入剖析了该类错误的产生机理，指出其根源在于训练初期模型性能不足，导致预设的评估操作点超出插值函数的定义域。并且针对此问题，本文提出了一种边界自适应插值策略（Boundary-Adaptive Interpolation Strategy, BAIS）。该策略通过动态感知模型性能边界与评估目标的相对位置，采用四阶段决策机制，将传统的被动错误捕获转变为主动的边界预判与处理。实验表明，该策略在模型性能极差或极优等边界场景下均能保持零错误率，较传统 try-except 方法在评估准确性和代码健壮性上均有显著提升。本研究为构建高可靠、全自动的深度学习训练与评估流程提供了重要的理论依据和技术支撑。</abstract><keywords>插值错误, 模型评估, 训练稳定性, 边界自适应, 深度学习, 人脸识别</keywords></article-meta></front><body/><back><ref-list><ref id="B1" content-type="article"><label>1</label><element-citation publication-type="journal"><p>[1] Schroff, F., Kalenichenko, D., &amp;amp; Philbin, J.. FaceNet: A Unified Embedding for Face Recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2015.
[2] Taigman, Y., Yang, M., Ranzato, M., &amp;amp; Wolf, L.. DeepFace: Closing the gap to human-level performance on face verification. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2014.
[3] 王伟 , 李明 , 张华 . 大规模分布式人脸识别训练中的稳定性优化研究 . 自动化学报 , 46(8), 1603-1612, 2020.
[4] Deng, J., Guo, J., &amp;amp; Zafeiriou, S.. ArcFace: Additive Angular Margin Loss for Deep Face Recognition. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019.
[5] 刘洋 , 陈云 . 面向安防应用的人脸识别系统性能评估方法 . 计算机研究与发展 , 56(4), 745-756, 2019.
[6] 李静 , 孙强 . 基于操作点约束的深度学习模型评估与选择 . 软件学报 , 32(5), 1321-1335, 2021.
[7] Virtanen, P., Gommers, R., Oliphant, T. E., et al. SciPy 1.0: fundamental algorithms for scientific computing in Python. Nature Methods, 17(3), 261-272, 2020.
[8] 张涛 , 赵磊 . 时间序列预测模型评估中的鲁棒性分析 . 计算机学报 , 41(11), 2421-2433, 2018.
[9] Goodfellow, I., Bengio, Y., &amp;amp; Courville, A.. Deep Learning. MIT Press, 2016.</p><pub-id pub-id-type="doi"/></element-citation></ref></ref-list></back></article>
