Romain Veltz

Why iter.tol is 5.4817113?? Otherwise, the behavior is actually normal. The matrix is badly conditioned. See: ```julia using BifurcationKit n = 10 A = Tridiagonal(-ones(n-1),2ones(n),-ones(n-1)) |> Array rhs = rand(n)...

It is somehow like the example you mention but I would like to avoid using a closure or `const` variables. > By the way, the example you show, the load...

This is what I do for `newton` ``` function _new_nlsolve_cache_RV(u,nls,feop) @assert Gridap.FESpaces.is_a_fe_function(u) x0 = get_free_values(u) op = Gridap.FESpaces.get_algebraic_operator(feop) f!(r,x) = Gridap.FESpaces.residual!(r,op,x) j!(j,x) = Gridap.FESpaces.jacobian!(j,op,x) fj!(r,j,x) = residual_and_jacobian!(r,j,op,x) f0, j0 =...

This is beautiful! I will play with it

Thank you! The docs of OrthogonalPolynomialsQuasi.jl are quite small. How do you encode BC, nonlinear terms... by hand?

Hi, I am not sure this is the right place but I follow your suggestion of using `OrthogonalPolynomialsQuasi.jl`. I guess I can do a Laplacian on (-1,1) with Dirichlet BC...

Wow!! Thank a lot for the detailed comment... I think I can change the tutorial on the Chan problem now by featuring ClassicalOrthogonalPolynomials.jl in 1D. > Though as I'm sure...

OK, I have this working: ```julia using Revise using BifurcationKit, LinearAlgebra, Plots, Setfield, Parameters, ClassicalOrthogonalPolynomials, ForwardDiff N(x; a = 0.5, b = 0.01) = 1 + (x + a*x^2)/(1 +...

I am not sure of how I handle the nonlinear term, do you agree with this?

Indeed...