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 2608.27122cs.LG

有向图学习的锥扩展瑞利商:极小极大谱认证、灵敏度与自适应控制

Cone Rayleigh Levels: Finite Perturbations and Certified Control

  • Institute of Mathematics with Computing Centre, Ufa Federal Research Centre of the Russian Academy of Sciences(俄罗斯科学院乌拉尔联邦研究中心数学与计算中心研究所)

机构由 AI 辅助整理,请以论文原文为准。

Yavdat Sh. Il'yasov, Nur F. Valeev

AI总结:

该研究针对有向图学习的非对称传播算子,基于锥瑞利框架开发了面向学习的谱认证、灵敏度分析与自适应控制方法,在有向Cora网络实验中可降低特定谱水平且不影响测试准确率。

AI中文摘要:

有向图学习自然会产生可训练的非对称传播算子,具有不同的左右谱结构。基于广义铅笔的双侧锥瑞利框架 \\( B_\theta-\lambda G \\),我们开发了一种面向学习的方法,用于谱认证、灵敏度分析和控制,无需对称性、非负性或锥保持性。在正卦限设置中,可计算的上下锥界为特定锥水平提供了后验包围,而平滑的软最小/最大代理保留了严格的单侧界,具有明确的近似误差,并且相对于可训练参数保持可微性。对于简单内部水平,左右模式满足 \\( D\lambda_C(B)[H]=v_C^T H u_C \\),在规定的扰动预算下产生由图支持的一阶最优干预,并激发自适应谱控制。数值实验证明了该方法在锥保持算子之外以及有向学习设置中的适用性。带符号的非对称扰动揭示了从内部特征对到边界互补准特征对的转变,包括非谱锥水平,而受控实验表明,对称化可去除仅由边方向携带的预测信息。在有向Cora引用网络上,在累积边权重减少预算为0.5%的情况下,左右灵敏度的自适应重新计算将特定谱水平降低了约21.5%,在所考虑的训练模型和数据划分中,未观察到测试准确率的变化。

英文摘要:

We study the reuse of positive trial profiles under finite perturbations of nonsymmetric matrix pencils B-λG. The lower and upper cone Rayleigh levels need not coincide and are defined without requiring positive eigenvectors. In the positive orthant with positive diagonal $G$, we derive computable perturbation bounds and determine the exact worst-case trial gap over prescribed independent entrywise perturbation classes. This yields the largest uniform radius meeting a given trial-gap tolerance for fixed profiles, certifying trial-value accuracy without recomputation. Experiments on a nonnegative operator and five signed matrices compare sufficient and optimal radii, directional thresholds, and cold- and warm-start recomputation. Several signed cases exhibit severely limited uniform profile reuse despite the optimality of the radius. Exact rational checks verify the reported bounds and worst-case constructions for stored numerical inputs. A finite-budget linear program and differentiable constraints illustrate applications to verified control and learning.

补充信息

↑