Changes between Version 1 and Version 2 of ArpackWrapper
- Timestamp:
- 11/21/06 11:01:37 (7 years ago)
Legend:
- Unmodified
- Added
- Removed
- Modified
-
ArpackWrapper
v1 v2 62 62 To solve a shifted-spectrum problem, ARPACK needs us to solve (A-sigma*B)*x = B*v for x, as well as perform B*v seperately. Practically this means we need to solve (A - sigma*B) using linalg.splu(). Here I'm assuming a generalised system. To solve the normal eigenproblem, set B = I (identity matrix). 63 63 64 == Advanced Python Inte face ==64 == Advanced Python Interface == 65 65 66 66 Sometimes you may want to use a user specified matrix-vector product or solver. E.g. the elements for the matrix vector product is calculated on the fly to save memory, or iterative solution (potentially using some specialised preconditioner) for shift-invert mode. The advanced interface could also provide more control over the type of ARPACK solver to use. My personal ARPACK python driver function looks like this:
