I try to reproduce the experiments (example 4 in official KAN). With official KAN, I get the results as below (Ground-truth is at the top, and the predication is at the bottom):
But with the efficient-kan, I get the results as below:
It shows that previous peak will be higher when learning new peak.
The official model is create by: "model = KAN(width=[1, 1], grid=200, k=3, noise_scale=0.1, bias_trainable=False, sp_trainable=False, sb_trainable=False)"
The efficient-kan model is created by: "model = KAN([1, 1], grid_size=200)"
It seems to be the same except for "bias_trainable=False, sp_trainable=False, sb_trainable=False".
I try to reproduce the experiments (example 4 in official KAN). With official KAN, I get the results as below (Ground-truth is at the top, and the predication is at the bottom):
But with the efficient-kan, I get the results as below:
It shows that previous peak will be higher when learning new peak. The official model is create by: "model = KAN(width=[1, 1], grid=200, k=3, noise_scale=0.1, bias_trainable=False, sp_trainable=False, sb_trainable=False)" The efficient-kan model is created by: "model = KAN([1, 1], grid_size=200)" It seems to be the same except for "bias_trainable=False, sp_trainable=False, sb_trainable=False".