ASCIIJK

> Regardless of computaion, do KAN at present able perform continual learning on 2D input data (or its flattened 1d vector features)? if not what is the main challenge KAN...

> > Regardless of computaion, do KAN at present able perform continual learning on 2D input data (or its flattened 1d vector features)? if not what is the main challenge...

> As I hypothesized in [the efficient-kan repo](https://github.com/Blealtan/efficient-kan/issues/28#issuecomment-2123449210), it seems KAN cannot do continual learning in more than 1 dimension if the output depends on more than 1 of them,...

Is it the reason for little effectness of CIL (class incremental learning)? I have tried to replace the fc layer directly with KAN layer on image classification task (CIFAR-100 B50inc5)....

> Anyone else found that KAN's effectiveness against catastrophic forgetting works only on 1D tasks? Since the spline locality applies to each dimension independently, it can't isolate more complex patterns...

> > Is it the reason for little effectness of CIL (class incremental learning)? I have tried to replace the fc layer directly with KAN layer on image classification task...

> > > > Is it the reason for little effectness of CIL (class incremental learning)? I have tried to replace the fc layer directly with KAN layer on image...