[R] SineKAN: Kolmogorov-Arnold Networks Using Sinusoidal Activation Functions

Excerpt

I couldn't sleep because I couldn't stop wondering if anyone had tried using sinusoids instead of B-splines as activation in a KAN, and fortunately/unfortunately that was already the case. I could not find it posted here, so I though I would share in the hope of some insightful discussion. Arxiv: https://arxiv.org/abs/2407.04149 Github repo: https://github.com/ereinha/SineKAN Also what appears to be a peer-reviewed "official" publication here: https://www.mdpi.com/2227-7390/13/19/3157 submitted