Change FFT to Real to Complex FFT

[tknopp]

Closes #85. This corresponds to the orginal FNO architecture and as outlined in #85, using the ordinary FFT will create different results. I would even say wrong results although the network might be able to correct this.

[codecov]

Codecov Report

Merging #87 (fed68b8) into main (60b7726) will increase coverage by 0.27%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main      #87      +/-   ##
==========================================
+ Coverage   90.17%   90.44%   +0.27%     
==========================================
  Files          10       10              
  Lines         173      178       +5     
==========================================
+ Hits          156      161       +5     
  Misses         17       17              

Impacted Files	Coverage Δ
src/Transform/chebyshev_transform.jl	100.00% <100.00%> (ø)
src/Transform/fourier_transform.jl	100.00% <100.00%> (ø)
src/operator_kernel.jl	100.00% <100.00%> (ø)

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[tknopp]

@yuehhua @ChrisRackauckas: What's your take on this?

[tknopp]

@foldfelis could you have a look? I am not sure who is maintaining this package. I developed a package based on NeuralOperators.jl and require the rfft patch.

[yuehhua]

Could you retain the coding style as usual?

[tknopp]

Could you retain the coding style as usual?

Sure, but could you give me a hint where this is not done? I tried to see it but it seems the coding style matches.

[yuehhua]

A simple way to format the coding style:
Change your directory to the NeuralOperators.jl folder and run the following code:

using JuliaFormatter
format(".", verbose=true)

It will format all the code in current folder for you.

[yuehhua]

It looks good to me.

[tknopp]

Thanks!