We want to do the code to support #187
It is not all that obvious what to pass into it to get the b^B. I think there are several interface approaches, but ultimately we need to be careful that we
(1) we build it by pieces as we do with the others. that is maybe the following would work?
bc = (Absorbing(), Reflecting())
b = [2.0; 0.0]
b_bc = r * L₀affine(x̄, bc, b) - μ*L₁₋affine(x̄, bc, b) - σ^2 / 2 * L₂affine(x̄, bc, b)
Then you can use this as the fudge when doing the linear solve
v = L_bc \ (f + b_bc)
(where I think we can just drop the L₀affine(x̄, bc) since it would be 0 ??? could be wrong, though).
This would require carefully checking that the math composes well.
There are other options as well that could be done. For example, the L₁₋bc(x̄, bc, b) is an overload of this for affine boundary conditions, and it returns a tuple:
In order to support solving models with non-homogeneous boundary conditions, we should see if it is feasible to make
[x] Add L₁₋affine etc (Resolved by 453fe99)
[x] Implement extrapolatetoboundary accordingly (Resolved by 74cf432)
[x] Update unit tests for extrapolatetoboundary (Resolved by 4acfdf2)
[x] Add unit tests for L₁₋affine etc (Resolved by 3dd2da6)
[x] Expand (205) to include the various versions of this for L_1+, L_1-, L_2, and L_0... Do these compose in a simple way (as we would need to automate? (See Laffine)
[x] Add package docs (Resolved by 07be253)
(the reset of the tasks related to example docs moved to a separate issue #193 )
After #187
We want to do the code to support #187 It is not all that obvious what to pass into it to get the
b^B
. I think there are several interface approaches, but ultimately we need to be careful that we (1) we build it by pieces as we do with the others. that is maybe the following would work?Then you can use this as the fudge when doing the linear solve
(where I think we can just drop the
L₀affine(x̄, bc)
since it would be0
??? could be wrong, though).This would require carefully checking that the math composes well.
There are other options as well that could be done. For example, the
L₁₋bc(x̄, bc, b)
is an overload of this for affine boundary conditions, and it returns a tuple:In order to support solving models with non-homogeneous boundary conditions, we should see if it is feasible to make
L₁₋affine
etc (Resolved by 453fe99)extrapolatetoboundary
accordingly (Resolved by 74cf432)extrapolatetoboundary
(Resolved by 4acfdf2)L₁₋affine
etc (Resolved by 3dd2da6)Laffine
)(the reset of the tasks related to example docs moved to a separate issue #193 )