home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.math
- Path: sparky!uunet!centerline!noc.near.net!news.cs.brandeis.edu!binah.cc.brandeis.edu!PALAIS
- From: palais@binah.cc.brandeis.edu
- Subject: Re: HELP. TEST FOR EXTREMA IN f(x,y,z)
- Message-ID: <1992Jul30.144053.28402@news.cs.brandeis.edu>
- Sender: news@news.cs.brandeis.edu (USENET News System)
- Reply-To: palais@binah.cc.brandeis.edu
- Organization: Brandeis University
- References: <1992Jul29.010542.8650@eng.ufl.edu>,<a_rubin.712451081@dn66>
- Date: Thu, 30 Jul 1992 14:40:53 GMT
- Lines: 21
-
- Armando Barre asked
-
- >My question is :
-
- >IS THERE SUCH A TEST FOR EXTREMA IN A FUNCTION OF 3 VARIABLES :f(x,y,z) ?
-
- Actually there is; someone else will have to provide a reference, but:
-
- (i) f(a,b) is a local maximum if the Hessian is negative definite
- (ii) f(a,b) is a local minimum if the Hessian is positive definite
- (iii) f(a,b) is a local minimum if the Hessian is indefinite.
-
- Arthur Rubin answered:
-
- I think Arthur meant to say saddle point for case (iii) (but one must
- assume that the Hessian matrix is non-singular). The same result is true
- in any dimension (and even in Hilbert space). The best reference is to any
- proof of the "Morse-Lemma", which says that after a change of variables the
- function locally is equal to a constant (its value at the critical point)
- plus the quadratic function given bythe Hessian. (See Lang's "Differentiable
- Manifolds" for the Hilbert space proof.)
-