Get the latest tech news

Convolutional-KANs


This project extends the idea of the innovative architecture of Kolmogorov-Arnold Networks (KAN) to the Convolutional Layers, changing the classic linear transformation of the convolution to learna...

Based on a 28x28 MNIST dataset, we can observe that the KANConv & MLP model achieves acceptable accuracy compared to the ConvNet (Big). Also the KKAN achieved 0.04 less Accuracy than ConvNet Medium, with almst half the parameter count (94k vs 157k), which shows the potential of the architecture. We are confident that as we increase the complexity of the models and the datasets we will see a significant improvement in the performance of the KAN Convolutional Networks.

Get the Android app

Or read this on Hacker News

Read more on:

Photo of KANs

KANs