I have k
triangular matrices of size d x d
. For each one, I want to solve for n
vectors (of dimension d
).
As far as I know, numpy.linalg.solve
allows me, using broadcasting, to perform all solves simultaneously. However, it uses Fortran's GESV
and cannot utilize the triangular structure.
On the other hand, scipy.linalg.solve_triangular
seems to be able to handle triangular matrices, but not support multiple simultaneous solves.
Is there a way to achieve both?
question from:
https://stackoverflow.com/questions/65841817/broadcast-solve-for-triangular-matrix 与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…