Lipschitz constant in FGP_TV for 3D

[epapoutsellis]

Hi @dkazanc

Why L = 26 here for the 3D case? I think it should be L=sqrt(12).

[dkazanc]

Hi, thanks this could be an error:
is it also sqrt(12) for 2D case here ?

[epapoutsellis]

No, 2D case is np.sqrt(8). I think (haven't check for different bc, it should be true) it should be
sqrt( 2^2 + 2^2 + ... number of dimensions + 2^2)

[dkazanc]

I see, so 2D: multip = (1.0f/(sqrt(8.0)*lambda)) and for 3D: multip = (1.0f/(sqrt(12.0)*lambda)) ?

[epapoutsellis]

Yes but without the square roots