@TSGut @dlfivefifty @MikaelSlevinsky I am witnessing instability in the lowering/raising when considering Semiclassical Jacobi with a large c parameter.
E.g. the following works:
ρ=0.02; t=inv(1-ρ^2);c = 5; SemiclassicalJacobi(t, 1, 1, c) \ SemiclassicalJacobi(t, 0, 1, c)
But the following fails (error given at the end):
ρ=0.02; t=inv(1-ρ^2);c = 100; SemiclassicalJacobi(t, 1, 1, c) \ SemiclassicalJacobi(t, 0, 1, c)
The reason I care is because I use these raising/lowering for the raising/lowering in annuli OPs. I have 3 questions
1) Should we fix this? Or..
2) Are we planning on changing the whole interface for the raising/lowering and I should just wait, Or..
3) Should I be implementing the raising/lowering of the annuli OPs directly (via QR/Cholesky) and not be going via the semiclassical Jacobi?
Simply switching to BigFloat currently has its own issues. But maybe 4.
@TSGut @dlfivefifty @MikaelSlevinsky I am witnessing instability in the lowering/raising when considering Semiclassical Jacobi with a large c parameter.
E.g. the following works:
ρ=0.02; t=inv(1-ρ^2);c = 5; SemiclassicalJacobi(t, 1, 1, c) \ SemiclassicalJacobi(t, 0, 1, c)
But the following fails (error given at the end):
ρ=0.02; t=inv(1-ρ^2);c = 100; SemiclassicalJacobi(t, 1, 1, c) \ SemiclassicalJacobi(t, 0, 1, c)
The reason I care is because I use these raising/lowering for the raising/lowering in annuli OPs. I have 3 questions1) Should we fix this? Or.. 2) Are we planning on changing the whole interface for the raising/lowering and I should just wait, Or.. 3) Should I be implementing the raising/lowering of the annuli OPs directly (via QR/Cholesky) and not be going via the semiclassical Jacobi?
Simply switching to
BigFloat
currently has its own issues. But maybe 4.4) Get the
BigFloat
implementation working.