SineKAN Replaces B-Splines with Sine Functions for Faster Inference

burkov · x · 2026-08-21

A paper from the University of Alabama introduces SineKAN, a KAN architecture that swaps heavy B-spline activations for re-weighted sine functions. It delivers substantial speedups across layer sizes and depths while matching standard NN accuracy on vision tasks.

Original post →

More from Research

Research channel →