home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!psinntp!kepler1!andrew
- From: andrew@rentec.com (Andrew Mullhaupt)
- Newsgroups: comp.lang.apl
- Subject: Re: Matrix Inverse in APL2
- Message-ID: <1474@kepler1.rentec.com>
- Date: 12 Jan 93 20:25:45 GMT
- References: <C0puI4.Mv6@undergrad.math.waterloo.edu>
- Organization: Renaissance Technologies Corp., Setauket, NY.
- Lines: 36
-
- In article <C0puI4.Mv6@undergrad.math.waterloo.edu> bdcrafto@cayley.uwaterloo.ca (Brad Crafton) writes:
- >HELP!! For a project I'm doing for work I require a routine to take the
- >Inverse of a very large matrix (in the order of 1000x1000). The
- >Matrix Inverse function tends to bomb out at even a 10x10 matrix.
-
- That doesn't sound like APL2's fault. APL2 uses Householder QR with
- column pivoting, so you may have a singular matrix in the 10x10 case,
- in which case you might be best served by: 1. thinking about the matrix
- to make sure it should have an inverse, 2. possibly using SVD to compute
- the pseudo-inverse.
-
- >I am assuming that some sort of Gaussian Elimination routine is
- >needed. Is there somewhere where I could find a pre-written routine?
-
- There are lots of places, but you should not so quickly assume that
- Gaussian elimination is a good idea.
-
- >I realize that for a matrix so large, it would probably take a LONG
- >time to run. But could anyone guestimate what kind of time I'm
- >looking at? Minutes? Hours? If it means anything, most of the
- >entries in the matrix will be zero.
-
- You should probably take advantage of the large number of zeros, (in the
- jargon you may have a 'sparse' matrix). The best thing there is to
- skip using APL at all and find a routine which may be suited to your
- particular matrix which can perhaps be used from within APL2. In order
- to get more out of the Internet, you should probably post a non-APL
- specific query with a reasonable description of your matrix problem to
- sci.math.num-analysis. Two books you should get some handle on, although
- you may need some elementary numerical analysis books before you can
- get very far, are Golub and van Loan's _Matrix Computations_ and Hanson
- and Lawson's _Solving Least Squares Problems_, although Golub and van
- Loan mostly supercedes Hanson and Lawson.
-
- Later,
- Andrew Mullhaupt
-