home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!wupost!usc!sol.ctr.columbia.edu!eff!news.byu.edu!news.mtholyoke.edu!nic.umass.edu!dime!bradtke
- From: bradtke@greed.cs.umass.edu
- Newsgroups: sci.math
- Subject: A matrix anti-derivative puzzle
- Message-ID: <52675@dime.cs.umass.edu>
- Date: 29 Aug 92 13:44:54 GMT
- Sender: news@dime.cs.umass.edu
- Reply-To: bradtke@greed.cs.umass.edu ()
- Distribution: usa
- Organization: University of Massachusetts, Amherst
- Lines: 32
-
-
- I have a problem involving the anti-derivative of a matrix function.
- I would like to find a function F such that
-
- dF/dw = - G'w
-
- or
-
- dF/dw = (G - G')w,
-
- where G is a nonsymmetric matrix, and w is a vector.
- The quadratic form F = w'Gw gives, of course, dF/dw = (G+G')w.
-
- Although if hasn't helped me any, the following may help you.
- The particular G I am concerned with is of the
- form
- G = H' (I-sP) Y H, where H is of dimensions (n by m), s is a scalar between
- 0 and 1, P is the transition probability matrix of a Markov chain, and
- Y = diag(y), where y is the vector of limiting state residence
- probabilities of the ergodic Markov chain.
-
- Thanks for any time you may put into this. Even if you only get a few
- partial results, I'm interested in seeing those. Maybe putting those
- together will achieve a complete answer.
-
- I'll share the results will all that are interested.
-
- Steve
-
- bradtke@cs.umass.edu
-
-
-