NTK引导的隐式神经教学
NTK-Guided Implicit Neural Teaching
对做隐式神经表示（图像拟合、3D重建、NeRF类场）的研究者，可直接复用其基于NTK增广梯度范数的坐标采样打分准则，替换均匀/误差驱动采样以在几乎减半的训练时间内达到同等重建质量。
原文依据 NINT scores examples by the norm of their NTK-augmented loss gradients
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论隐式神经表示（INR）通过多层感知机（MLP）参数化连续信号，可为图像、音频和三维重建等任务提供紧凑且与分辨率无关的建模。然而，拟合高分辨率信号需要在数百万个坐标上进行优化，带来难以承受的计算开销。为此，我们提出NTK引导的隐式神经教学（NINT），通过动态选择能最大化全局函数更新的坐标来加速训练。借助神经正切核（NTK），NINT以样本经NTK增广后的损失梯度范数对其打分，从而同时捕捉拟合误差与异质杠杆效应（自影响与坐标间耦合）。这种双重考量使收敛速度快于现有方法。通过大量实验，我们证明NINT在保持甚至提升表示质量的同时，将训练时间显著缩短近一半，在近期基于采样的加速策略中达到最优水平。
英文原文摘要 1062 字符
Implicit Neural Representations (INRs) parameterize continuous signals via multilayer perceptrons (MLPs), enabling compact, resolution-independent modeling for tasks like image, audio, and 3D reconstruction. However, fitting high-resolution signals demands optimizing over millions of coordinates, incurring prohibitive computational costs. To address it, we propose NTK-Guided Implicit Neural Teaching (NINT), which accelerates training by dynamically selecting coordinates that maximize global functional updates. Leveraging the Neural Tangent Kernel (NTK), NINT scores examples by the norm of their NTK-augmented loss gradients, capturing both fitting errors and heterogeneous leverage (self-influence and cross-coordinate coupling). This dual consideration enables faster convergence compared to existing methods. Through extensive experiments, we demonstrate that NINT significantly reduces training time by nearly half while maintaining or improving representation quality, establishing state-of-the-art acceleration among recent sampling-based strategies.