home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.math.num-analysis
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!sol.ctr.columbia.edu!usenet.ucs.indiana.edu!bronze.ucs.indiana.edu!dhart
- From: dhart@bronze.ucs.indiana.edu (dave hart)
- Subject: Re: Chebyshev vs. Least Squares Polynomials
- Message-ID: <BzEyyF.GE0@usenet.ucs.indiana.edu>
- Sender: news@usenet.ucs.indiana.edu (USENET News System)
- Nntp-Posting-Host: bronze.ucs.indiana.edu
- Organization: Indiana University
- References: <1992Dec2.170529.24897@news2.cis.umn.edu>
- Date: Thu, 17 Dec 1992 17:23:02 GMT
- Lines: 25
-
- In article <1992Dec2.170529.24897@news2.cis.umn.edu>
- pires@milli.cs.umn.edu (Luiz S. Pires) writes:
- >
- > Please forgive me if these are trivial questions. I am not a mathematician.
- >
- > When approximating a function with a polynomial, should one use the
- > least squares approximating polynomial or a Chebyshev polynomial?
- >
- > Does L.S. minimize overall error and Chebyshev minimize maximum error?
- >
- > Which is more commonly used in approximating transcendental functions?
- >
- > Also, when are other methods (eg Rational Chebyshev approximation) used?
- > Any good references?
- >
- The idea of "least squares" is the same as orthogonal projection--
- "the shortest distance between a point and a line..." This is tied up
- with geometry, ie a metric [inner product]. The Chebyshev polynomials
- form an orthonormal basis [q.v.] for the L^2 metric; other functions
- form orthonormal bases for other metrics.
-
- I cannot guess what you mean by "least squares approximating
- polynomial," but approximation implies a notion of distance; best
- approximation(s) follow from that.
-
-