search

AntonioTepsich / 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 learnable non linear activations in each pixel.
MIT License
779 stars 76 forks source link

Great! How about the train/infer time? #2

Closed yuedajiong closed 5 months ago

yuedajiong commented 6 months ago

as title

and if we want better precision against to big conv, how about the parameters increasing?

KKAN means CKAN,...,CKAN,...,KAN,... CKAN+MLP means CKAN,...,CKAN,...,MLP,... right?

thanks!

XiangboGaoBarry commented 6 months ago

Hi, here I implement ConvKAN with different activation formulations with their corresponding inference time. https://github.com/XiangboGaoBarry/ConvKAN-Zoo

yuedajiong commented 6 months ago

@XiangboGaoBarry
Greaaaaaaaaaaaaaaaaaaaaaaaaaaaat!

看到了，牛逼，细致。 已经下载了在看。 （刚看到了一个repo: https://github.com/cloneofsimo/imagenet.int8 )我想从最最最原始的一步开始，走到我git主页搞的那个3d task)

除了祝你发大财，好像其他的赞美都不够!!!!!!!!!!!!!!!!!