Trying Kolmogorov-Arnold Networks in Practice (cprimozic.net)

Created on 2025-05-03T05:13:26-05:00

Return to the Index

This card pertains to a resource available on the internet.

This card can also be read via Gemini.

KAN networks are more difficult to get good performance out of compared to traditional neural networks.

KAN networks are not "silver bullets" of AI models but have orthogonal usefulness: more interpretable, or fitting given degrees of mathematical precision by tuning the learnable activation functions.

No matter what I did, the most simple neural network was still outperforming the fanciest KAN-based model I tried.