home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.arch:10986 comp.lang.misc:3826
- Newsgroups: comp.arch,comp.lang.misc
- Path: sparky!uunet!spool.mu.edu!news.nd.edu!mentor.cc.purdue.edu!hrubin
- From: hrubin@mentor.cc.purdue.edu (Herman Rubin)
- Subject: Re: Hardware Support for Numeric Algorithms
- Message-ID: <By2KM8.Fxp@mentor.cc.purdue.edu>
- Organization: Purdue University Statistics Department
- References: <722061187@sheol.UUCP> <BxxGKp.81A@mentor.cc.purdue.edu> <1992Nov20.054603.28338@technix.mn.org>
- Date: Sat, 21 Nov 1992 14:08:31 GMT
- Lines: 41
-
- In article <1992Nov20.054603.28338@technix.mn.org> bret@technix.mn.org (Bret Indrelee) writes:
- >In article <BxxGKp.81A@mentor.cc.purdue.edu> hrubin@pop.stat.purdue.edu (Herman Rubin) writes:
- >>The statistician will need numerical analysis jargon. That computer
- >>jargon diverges from this is essentially a move on the part of the
- >>computer language designers to distance themselves from mathematics.
- >>They have been too succesful in this.
-
- >You could say the same thing in terms of Electrical Engineering or
- >Chemistry moving too far from Physics.
-
- ......................
-
- >The focus of a computer programmers work is much different than
- >a mathematicians. The mathematician can prove that an algorithem
- >works, and be done with it. A computer programmer that doesn't
- >worry about how many steps the algorithem takes (how long to
- >complete a run) is going to get into lots of trouble.
-
- >Different focus. Different jargon. Each can ignore different details.
-
- The two aspects I have been arguing are the notation to communicate,
- and the speed of the object code. This is also what those using the
- computers in number theory are arguing. The physicist does not attempt
- to use notation known to confuse the chemist or electrical engineer.
- The numerical analyst does not tell the engineer to state it in a new
- "clean" format or go without the services. When I consult on statistical
- problems, the problem is not statistical notation, but that the client
- does not know that more SUBJECT MATTER assumptions are needed to solve
- the problem, and that the numerical answers depend on those assumptions
- in a non-trivial manner. Also, often there is a partial analysis, with
- further work needed on steps, even on the decision of what data to collect.
-
- It is precisely this approach which the programming people seem unwilling
- to take. I do not attempt to force the solution method into a small
- algorithmic framework, and often invent algorithms for the particular
- problem.
- --
- Herman Rubin, Dept. of Statistics, Purdue Univ., West Lafayette IN47907-1399
- Phone: (317)494-6054
- hrubin@snap.stat.purdue.edu (Internet, bitnet)
- {purdue,pur-ee}!snap.stat!hrubin(UUCP)
-