This addresses #77, which details a number of at the time unknown issues with the accuracy of PyExaFMM. I can report that there were 3 main issues, which are addressed in this PR.
1) Calculation of the pseudoinverse of highly ill conditioned matrix
2) Random BS bugs (casting contribution from W List to int rather than float, thus rounding it to 0)
3) A catastrophic bug in the calculation of the V List for leaf nodes not discretised to maximum level, which was fixed here
This addresses #77, which details a number of at the time unknown issues with the accuracy of PyExaFMM. I can report that there were 3 main issues, which are addressed in this PR.
1) Calculation of the pseudoinverse of highly ill conditioned matrix 2) Random BS bugs (casting contribution from W List to int rather than float, thus rounding it to 0) 3) A catastrophic bug in the calculation of the V List for leaf nodes not discretised to maximum level, which was fixed here