From: <HIDDEN>
Newsgroups: comp.soft-sys.matlab
Subject: Re: Piecewiese linear approximation function with a minimal largest error dev.
Date: Wed, 2 Jan 2013 09:06:08 +0000 (UTC)
Organization: The MathWorks, Inc.
Lines: 8
Message-ID: <kc0ta0$ftf$>
References: <kbu8d3$gcr$> <kbv7d0$obt$> <kc0mm4$oem$> <kc0oak$dm$>
Reply-To: <HIDDEN>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
X-Trace: 1357117568 16303 (2 Jan 2013 09:06:08 GMT)
NNTP-Posting-Date: Wed, 2 Jan 2013 09:06:08 +0000 (UTC)
X-Newsreader: MATLAB Central Newsreader 1187260
Xref: comp.soft-sys.matlab:785749

"Bruno Luong" <b.luong@fogale.findmycountry> wrote in message <kc0oak$dm$>...
> As I told earlier, if the knot are fixed (regardless they are is equidistant or not) , this problem is a first-degree spline-approximation (piece-wise linear) of l_infinity norm, and can be formulated as linear programming.
> Bruno
- - - - - - - - -
  Bruno, we seem to be looking at Deyan's problem from two different aspects.  You are saying that given an L_infinity norm the problem can be solved as a linear programming problem and you may well be right.  I have been saying that its primary difficulty lies in calculating this L_infinity norm since it involves not just the differences at the segment endpoints which would involve a finite number of variables but the differences occurring between these endpoints where an infinite continuum of points are involved.  That is why I spoke of the inverse of derivatives or numerous intermediate points.  Please correct me if I am mistaken.

Roger Stafford