home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!europa.asd.contel.com!darwin.sura.net!zaphod.mps.ohio-state.edu!magnus.acs.ohio-state.edu!usenet.ins.cwru.edu!agate!mickey.cs.berkeley.edu!bsmith
- From: bsmith@mickey.NoSubdomain.NoDomain (Brian Smith)
- Newsgroups: comp.compression
- Subject: Re: Math Question
- Date: 16 Nov 1992 03:53:00 GMT
- Organization: University of California, Berkeley
- Lines: 81
- Sender: bsmith@mickey (Brian Smith)
- Distribution: world
- Message-ID: <1e75usINN31v@agate.berkeley.edu>
- References: <24005@hacgate.SCG.HAC.COM>
- NNTP-Posting-Host: mickey.cs.berkeley.edu
-
- In article <24005@hacgate.SCG.HAC.COM>, abirenbo@rigel.cel.scg.hac.com (Aaron Birenboim) writes:
- |>
- |> While working on an optimization problem related to compression
- |> I bumped into the following nasty little problem :
- |>
- |> evaluate:
- |>
- |> { [(i+1)q }
- |> d { [ }
- |> --- { Integral [ (x - (i + 0.5) q)^2 * exp(-c * x) dx }
- |> dq { [ }
- |> { [i*q }
- |>
- |> I have evaluated the integral..... and it is terribly nasty. Even if I
- |> didn't goof on the evaluation, i would surely goof on the differentiation.
- |>
- |> c and i are constants. I'm trying to minimize error for q.
- |>
- |> I seem to remember some terribly nasty formulas for handling the
- |> differentiation before evaluating any integral.
- |>
- |> I think it involved the bounds, derivitives of the bounds, the integrand
- |> and the derivitave of the integrand.
- |>
- |> My memory comes from a dim recolection of some stuff covered
- |> by Dr. Sawchuk at USC (southern cal) in a class on random variables
- |> based on the book by papoulis. I cannot find it mentioned in papoulis...
- |> i think is was ONLY in Sawchuk's notes (which i cannot find).
- |>
- |> any advise, help, or refrences are welcome.
- |>
- |> If i knew the NAME of this theorem, I might be able to look it up.
- |>
- |> thanks in advance,
- |>
- |> --
- |> Aaron Birenboim |
- |> birenb@hac2arpa.hac.com |
- |> |
- |> H (303) 871-8271 |
-
- I believe the following is correct (this is the result of about 2 minutes
- of calculation, but it seems to work). I decided to post it so other people
- could check it, since I'm not perfectly confident in the answer (my calculus
- was a long time ago!).
-
- Theroem:
-
- If f(x) is a function that is analytically integrable, then
-
- { [y }
- d { [ }
- --- { Integral [ f(x) dx } = f(y)
- dy { [ }
- { [0 }
-
- Proof:
-
- Let g(x) be the indefinate integral of f(x). Then, by the fundamental theorem
- of integral calculus, g'(x) = f(x). Now, the definate intergral from 0 to y
- of f(x) is g(y) - g(0). The derivative of this expression wrt y is g'(y) which
- is f(y). QED
-
- Given this, and a few reductions, the expression
-
- { [(i+1)q }
- d { [ }
- --- { Integral [ (x - (i + 0.5) q)^2 * exp(-c * x) dx }
- dq { [ }
- { [i*q }
-
- evaluates to
-
- q^2 * exp(-c * q) * [ ({(i+1) - (i+0.5)}^2) - ({i - (i+0.5)}^2) ]
-
- which, if I didn't make any algebra errors, is zero.
-
- -----
- Brian C. Smith arpa: bsmith@cs.Berkeley.EDU
- University of California, Berkeley uucp: uunet!ucbvax!postgres!bsmith
- Computer Sciences Department phone: (510)642-9585
-