Apart from the Fourier Neural Operator, the other prominent architecture for operator approximation is DeepONet. This should be implemented in a similar fashion as well.
Something like:
model = DeepONet(in_branch, in_trunk, out, latentspace, activation; biases...)
It won't make sense to distinguish between one singular layer and the entire architecture similar to FourierLayer. So this should be worked on after #22 is resolved.
Apart from the Fourier Neural Operator, the other prominent architecture for operator approximation is DeepONet. This should be implemented in a similar fashion as well.
Something like:
It won't make sense to distinguish between one singular layer and the entire architecture similar to
FourierLayer. So this should be worked on after #22 is resolved.