home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!olivea!charnel!rat!usc!howland.reston.ans.net!zaphod.mps.ohio-state.edu!darwin.sura.net!newsserver.jvnc.net!netnews.upenn.edu!netnews.cc.lehigh.edu!ns1.cc.lehigh.edu!fc03
- From: fc03@ns1.cc.lehigh.edu (Frederick W. Chapman)
- Newsgroups: sci.math.symbolic
- Subject: Re: Symbolic Fortran : Help Needed
- Message-ID: <1993Jan25.154419.75401@ns1.cc.lehigh.edu>
- Date: 25 Jan 93 15:44:19 GMT
- Organization: Lehigh University
- Lines: 52
-
- In article <1993Jan22.081540.1700@atri.curtin.edu.au>,
- thng@atri.curtin.edu.au (Ian Li-Jin Thng) writes:
-
- >G'day to all symbolic netters,
- >
- >I'm seeking assistance on anyone who has experience with or have
- >come across references regarding symbolic programming using Fortran
- >77 .
- >
- >I know that a common method of symbolic programming is to use
- >pointer structures which Fortran 77 do not support.
- >
- >I have at present a Fortran program which is numerically intensive
- >at the moment. I need to add a section to that program which is
- >unfortunately symbolically intensive. The requirements of the
- >symbolically intensive portion are as follows :
- >
- >Essentially to find out the symbolic expressions for the partial
- >derivatives [...]
-
- There is a package called JAKEF in the NETLIB collection which will
- perform "automatic differentiation" on scalar- and vector-valued
- functions of several variables, thereby yielding the gradient and
- Jacobian, respectively. The results are exact up to round-off error.
- (I will post an "Introduction to NETLIB" document separately to this
- news group.)
-
- For symbolic differentiation in FORTRAN (often desirable for
- optimization routines which need Jacobians and sometimes Hessians),
- you might want to search the literature on the topic "automatic
- differentiation", which refers to the process of taking a subroutine
- that computes a given differentiable function and automatically
- producing a second subroutine that computes the derivative of that
- function by an exact rather than approximate method. (There is a
- description of this in the Spring '92 issue of the Maple Technical
- Newsletter.) You might also want to query sci.math.num-analysis on
- the topic of "automatic differentiation".
-
- Good luck,
-
- Frederick W. Chapman
- Senior User Consultant
- Lehigh University Computing Center
-
- --
-
- o ------------------------------------------------------------------------- o
- | Frederick W. Chapman, User Services, Computing Center, Lehigh University |
- | Campus Phone: 8-3218 Preferred E-mail Address: fc03@Lehigh.Edu |
- o ------------------------------------------------------------------------- o
- | Ecstasy is transitory, but a theorem is forever! |
- o ------------------------------------------------------------------------- o
-