How to take derivative of complex functions like Weierstrass Elliptic function?

Specific Domains
1

I want to use Elliptic and modular functions as activation functions in neural networks. What packages should I use? I can’t install EllipticFunctions.jl because it returns error.
How could I apply chain rule to optimize them.

2

Unfortunately, the author of the EllipticFunctions.jl package (stla, Stéphane Laurent) died in August 2024 and the package has not been updated to work with the latest Julia versions.

1 Like
3

Sorry. I didn’t know that. Is there any other package that I could use?

4

What error are you seeing? It seems to work fine for me… It may not have been updated since 2024, but there shouldn’t be any updates required to work on newer Julia versions.

That said, we should probably move the repository to JuliaMath so it can receive bug fixes as needed.

2 Likes
5

The OP said that EllipticFunctions.jl cannot be installed. I didn’t test it.

6

Thanks. I tried to isolate EllipticFunctions.jl from Bukdu.jl and solved my problem.

julia> gradient(x -> real(thetaC(x, tau=0.5 + 1im)), 1.0)
(-0.8061618539015968,)

julia> gradient(x -> real(thetaC(x, tau=0.5 + 1im)), 0.5)
(-0.4433516470405768,)

I’m also greatful for Laurent’s incredible library.