Tag
论文提出 L-SplineFNO，一种将 FNO 主干与稀疏可学习张量积 B 样条传输耦合的内存高效神经算子，用于在非结构化网格上近似 PDE 解算子，在六个三维工程基准上实现高精度并可扩展至 4600 万节点，还支持零样本超分辨率和基于 Bernstein 凸包与 Taylor 模型的严格预测界。
FlashKAN proposes a method to accelerate Kolmogorov-Arnold Networks by replacing Cox-de Boor recursion with truncated power form for B-spline evaluation, providing a fused GPU implementation and an open-source package.
Introduces PI-Splines, a structured spline-based architecture for physics-informed learning that parametrizes unknown fields with trainable B-spline coefficients, providing compact support, analytical derivatives, and strong boundary condition enforcement, demonstrated as a competitive alternative to neural network-based methods.