-
Notifications
You must be signed in to change notification settings - Fork 1.2k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
How to support complex data type? #87
Comments
Hi, unfortunately pykan doesn't support complex numbers. However, you may try treating real and imaginary parts separately, so you end up feeding KAN with vectors of doubled length. pykan doesn't support GPU out-of-the-box, so I'd suggest debugging (with small-scale datasets) on cpus first. |
Is there any particular reason in your domain that you don't want to separate real and imaginary parts? I know in some applications people want to constrain to holomorphic functions where complex neural networks are favored. In other cases, I don't see a strong reason not to just treat real and imaginary separately and feed into real-valued neural networks (which are much more optimized than complex-valued ones). |
Something like Earth gravity fields required N summation of spherical harmonics. However, I currently do not figure out (1) how to add multiple spherical harmonics base functions as symbolic function. (2) how to apply complex numbers. |
@plyu3 thanks for the feedback. Currently it's not supported but sounds like something worth being included in the future. |
Hi. It's a great project.
I want to use KAN in radar signal processing domain. As you know that the radar signal is complex number. When I create a dataset with complex data and try to train KAN it report errors as below:
How to solve this problem? I prepare the development environment by using pip install -r requirements. I notice that the pytorch is CPUversion. What about I switch to GPU version Pytorch? Can it solve this problem?
The text was updated successfully, but these errors were encountered: