home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky sci.math.num-analysis:2676 sci.math.symbolic:2367
- Newsgroups: sci.math.num-analysis,sci.math.symbolic
- Path: sparky!uunet!mcsun!Germany.EU.net!infko!inews
- From: berling@infko.uni-koblenz.de
- Subject: decomposing sets of equations
- Message-ID: <1992Sep10.131427.22221@infko.uucp>
- Sender: inews@infko.uucp (inews)
- Organization: University of Koblenz, Germany
- Date: Thu, 10 Sep 1992 13:14:27 GMT
- Lines: 53
-
- Hello!
-
- I have some questions concerning methods for solving large sets
- of mostly nonlinear equations (it's a kind of constraint solving problem).
-
- I am seeking for methods to decompose a possible large set of
- equations into sets of smaller ones which have to be solved
- simultaneously in a certain order. Using a bipartite graph with
- vertices representing the equations and variables
- and undirected arcs representing
- the "belongs-to" relationship, one may use a maximum-matching algorithm
- to determine a partition into smaller sets. My question is, are there
- any other techniques, especially techniques which allow a better
- control of the quality of the generated partition (e.g. number of sets of
- more than one equation is minimal)?
-
- Having a set of equations which have to be solved simultaneously, I am
- further interested in methods for eliminating as much variables as
- possible before solving it using Newton-Raphson iteration. Having for
- example a set of 4 equations determining the variables a, b, c, d:
-
- 1. a = f (c, d)
- 2. b = g (a, c)
- 3. c = h (d)
- 4. d = i (b)
-
- the system is reducible to the equation:
-
- b = g (f (h (i (b)), i (b)), h (i (b)))
- or
- d = i (g (f (h (d), d), h (d)))
-
- which may be easier to solve using Newton-Raphson. Knowing wich equation
- has to be solved for which variable, the problem has probably something
- to do with finding the feedback vertices in strong components of
- directed graphs, a problem known to be NP-complete. Again, my question is
- are there any other techniques or heuristics which solve this problem
- or special cases perhaps more efficient?
-
- I will thank you for any hints!
-
- +----------------------------+-----------------------------------------+
- | | |
- | Roland Berling | Tel.: +49 261 9119 436 |
- | Universit"at Koblenz | Fax.: +49 261 9119 499 |
- | Institut f"ur Informatik | e-mail: berling@infko.uni-koblenz.de |
- | Rheinau 3-4 | |
- | D-5400 Koblenz | |
- | Germany | |
- | | |
- +----------------------------+-----------------------------------------+
-
-
-