mm-1014
===
Subject: Re: roots of quadratic matrix equation?
> Thus formally you have two solutions the same way you have
solutions for
> n = 1:
> Q = 0.5*B*A^(-1) +(-) ( 0.25*B*B - C*A^(-1) )^(1/2)
> But in this case you have a problem of finding the square
root of matrix
> (P*P = M => P = M^(1/2) ) which is as difficult as an
initial problem.
Not really, you can quickly roll your own recursive (sqrt)
solver,
P = (P + M*inv(P))/2
------------------------------------------------------
Applied Algorithms http://sdynamix.com
===
Subject: Help with Template Numerical Toolkit (TNT)
I have included the tnt.h and jama.h files in multiple 
files of
my code.
If i run the program with these files included only in one .C
it runs
perfectly.
But i cannot access it from the other .C.
If i include in both it gives me the following error
multiple definition of 
ÔTNT::Stopwatch:;Stopwatch(void)
I have not been able to figure out why.
Is therea possibility of a problem with my Make file.
Please help
Nachiket
===
Subject: Elliptic integrals-->complex arguments-->migraine
I am way over my head in the area but the basic problem
is how to compute a Elliptic integral of the second
kind with a complex argument?
E(w|m) = ?
where w = u + iv.
Ab.&Stegun mentions the case for E(phi+i chi alpha) but
this notation game with the ellipticals is giving me a
severe migraine.
17.4.12, p. 593.
===
Subject: Re: Elliptic integrals-->complex arguments-->migraine
> I am way over my head in the area but the basic problem is
how to
> compute a Elliptic integral of the second kind with a
complex argument?
> E(w|m) = ?
> where w = u + iv.
> Ab.&Stegun mentions the case for E(phi+i chi alpha) but
this notation
> game with the ellipticals is giving me a severe migraine.
> 17.4.12, p. 593.
looking at 17.2.1, m = sin^2alpha, w=phi + i chi.
Hope this helps,
john
===
Subject: Numerical Integrators for N-Body simulation of the
solar system
Im trying to select an appropriate numerical integration
scheme for
doing an NBody simulation of the solar system (planets,
moons, plus a
few significant asteroids and comets).
Ive been researching what numerical integration scheme 
would
be best to
use.
Several people who have done solar system simulations seem to
have gone
with a Bulirsch-Stoer scheme, or Everharts RADAU scheme.
However, among astrophysicists doing cutting edge simulations,
symmetrized Hermite schemes seem to be used almost
universally.
Meanwhile, NASAs ephemeris used an Adams-Bashforth scheme.
I have also heard a lot about symplectic methods, though I
dont know
the first thing about the theory, except that they preserve
various
properties of hamiltonians.
What is guiding all these choices? Is it a question of scale?
Suitability for special hardware? Or some other consideration?
Can anyone offer any suggestions on whats the best option
for a
relatively modest solar system simulation for graphical
display?
Mark Cummins.
CC: markcummins25@yahoo.co.uk
===
Subject: Re: Numerical Integrators for N-Body simulation of
the solar
system
> Im trying to select an appropriate numerical integration
scheme for
> doing an NBody simulation of the solar system (planets,
moons, plus a
> few significant asteroids and comets).
> Ive been researching what numerical integration scheme
would be best to
> use.
> Several people who have done solar system simulations seem
to have gone
> with a Bulirsch-Stoer scheme, or Everharts RADAU scheme.
> However, among astrophysicists doing cutting edge
simulations,
> symmetrized Hermite schemes seem to be used almost
universally.
> Meanwhile, NASAs ephemeris used an Adams-Bashforth 
scheme.
> I have also heard a lot about symplectic methods, though I
dont know
> the first thing about the theory, except that they preserve
various
> properties of hamiltonians.
> What is guiding all these choices? Is it a question of
scale?
> Suitability for special hardware? Or some other
consideration?
> Can anyone offer any suggestions on whats the best option
for a
> relatively modest solar system simulation for graphical
display?
> Mark Cummins.
To my knowledge the best choice for this kind of problem is
diva
available at http://mathalacarte.com. Depending on your use
of the code
you may be expected to pay for it. This is the code used for
these kind
of problems at the Jet Propulsion Laboratory. For smaller
problems the
RK code on my site might also be useful. I wouldnt even
bother testing
a Bulirsch_Stoer code. Diva allows you to skip the second
derivative
of a step, and this is likely to speed things up a little
when used on
the small bodies with elliptic orbits. This is a bit like a
Adams-Bashforth method. By all means use the option in the
code that
allows you to integrate second order equations directly.
Unless you are doing extremely long integrations, the
symplectic methods
will not be useful.
Fred
===
Subject: Re: Numerical Integrators for N-Body simulation of
the solar
system
See this link: http://www.midwestcs.com/xstar/
which provides a pretty enlightening primer on solving the
N-Body problem.
You can download an introductory text from
http://www.midwestcs.com/xstar/index.html in which the author
(Wayne
Schlitt) has much to say about sources of errors, numerical
schemes and
their ups and downs.
that used a 3rd order taylor series for most calculations
stating that
given the errors that come with many methods an excellent
balance between
accuracy and computational expense can be had. I had thought
it was on
the X-Star site, but I think I was mistaken.
> Im trying to select an appropriate numerical integration
scheme for
> doing an NBody simulation of the solar system (planets,
moons, plus a
> few significant asteroids and comets).
> Ive been researching what numerical integration scheme
would be best to
> use.
> Several people who have done solar system simulations seem
to have gone
> with a Bulirsch-Stoer scheme, or Everharts RADAU scheme.
> However, among astrophysicists doing cutting edge
simulations,
> symmetrized Hermite schemes seem to be used almost
universally.
> Meanwhile, NASAs ephemeris used an Adams-Bashforth 
scheme.
> I have also heard a lot about symplectic methods, though I
dont know
> the first thing about the theory, except that they preserve
various
> properties of hamiltonians.
> What is guiding all these choices? Is it a question of
scale?
> Suitability for special hardware? Or some other
consideration?
> Can anyone offer any suggestions on whats the best option
for a
> relatively modest solar system simulation for graphical
display?
> Mark Cummins.
===
Subject: Re: Numerical Integrators for N-Body simulation of
the solar
system
> Im trying to select an appropriate numerical integration
scheme for
> doing an NBody simulation of the solar system (planets,
moons, plus a
> few significant asteroids and comets).
> Ive been researching what numerical integration scheme
would be best to
> use.
> I have also heard a lot about symplectic methods, though I
dont know
> the first thing about the theory, except that they preserve
various
> properties of hamiltonians.
> What is guiding all these choices? Is it a question of
scale?
Yes, for long-term studies (millions of years) symplectic
methods
are essential. Other methods tend to lose energy even when
used on
a 2-body problem, which makes long-term predictions completely
useless. Or you need to work in special coordinate systems
which
exploit the near integrability of the solar system.
Arnold Neumaier
===
Subject: double integral question
Hi all,
Does anyone have experience with a type of integral like this:
int(int(P(k,x)*exp(-a*abs(x-x0))*P(m,x0),x=-1..1),x0=-1..1)
where P(k,x) is the k-th Legendre polynomial...
I need to evaluate a lot of these so Im looking for a fast
and robust
evaluation scheme. I could use numerical integration (and
splitting up the
integrals and intervals to get rid of the absolute value
discontinuity),
but that would be the least attractive solution from the
point of view of
efficiency and hence the final resort if all else 
fails...
Tips are very much appreciated...
gert
===
Subject: Re: double integral question
> Hi all,
> Does anyone have experience with a type of integral like
this:
> int(int(P(k,x)*exp(-a*abs(x-x0))*P(m,x0),x=-1..1),x0=-1..1)
> where P(k,x) is the k-th Legendre polynomial...
> I need to evaluate a lot of these so Im looking for a 
fast
and robust
> evaluation scheme. I could use numerical integration (and
splitting up
the
> integrals and intervals to get rid of the absolute value
discontinuity),
> but that would be the least attractive solution from the
point of view of
> efficiency and hence the final resort if all 
else fails...
> Tips are very much appreciated...
> gert
If you should decide to this numerically, then you want to
split this up
into two integrals in order to avoid the discontinuity. The
multidimensional quadrature code at http://mathalacarte.com
would be
hard to beat for this type of problem. That code has hooks in
it so
that you could specify the location of the discontinuity
without
actually breaking this up into two integrals. (However that
is what the
code will do internally in this case.) Depending on your use,
you may
be expected to pay for the software.
Fred
===
Subject: Re: double integral question
> Hi all,
> Does anyone have experience with a type of integral like
this:
> int(int(P(k,x)*exp(-a*abs(x-x0))*P(m,x0),x=-1..1),x0=-1..1)
> where P(k,x) is the k-th Legendre polynomial...
> I need to evaluate a lot of these so Im looking for a 
fast
and robust
> evaluation scheme. I could use numerical integration (and
splitting up
the
> integrals and intervals to get rid of the absolute value
discontinuity),
> but that would be the least attractive solution from the
point of view of
> efficiency and hence the final resort if all 
else fails...
> Tips are very much appreciated...
> gert
Well, manifestly, the matrix is symmetric and
positive-definite, so you
only have to do n*(n+1)/2 elements rather than n^2.
Second question is How big are k and m ? If you only need
low-order
polynomials, Gauss-Legendre integration is what you want. How
precise
do the results need to be?
You might want to consider the representation
exp( -a*|x-y| )
= (1/pi) * int_{-infty}^{infty} { du exp[ i*u*(x-y) ] / (u^2 +
a^2) }
Thus you do the integrals up to k = k_max
I_k (u) = int_{-1}^{+1} {dx exp(i*u*x) * P_k (x) }
and then evaluate the desired matrix elements
M_{k,m} = (1/pi) * int_{-infty}^{infty} { {I_k (u) * (I_m (u)
)^*
}
over { (u^2 + a^2) } } .
At least, thats how Id tackle it.
--
Julian V. Noble
Professor Emeritus of Physics
^^^^^^^^
http://galileo.phys.virginia.edu/~jvn/
Science knows only one commandment: contribute to science.
-- Bertolt Brecht, Galileo.
===
Subject: Re: Robust Parameter Estimation
I use lsqnonlin in Matlabs Optimization Toolbox. I have an
m-file,
say
fun.m, in which I implement my nonlinear function f(u;x)
given the set
of
model parameters,x, and the set of inputs, {u_i}. The output
of the model
is hat{y}_i, which is compared with experimental data, y_i.
The
differences are the residuals, which I try to minimize using
the cost
function J(e) given below.
> Hi
> I would like to know what Algorithm you use to solve this
problem by
> nonlinear least-squares.
> Algorithm, you only provide (u_i,y_i) and the nonlinear
function.
> Alien+
> My nonlinear least-squares cost if of the form:
> J(e) = sum_{i=1}^{N} [y_i - f(u_i, x)]^2, (Eq. 1)
> where f= f(u,x) is the model output for an input u and x is
the
vector
> (in
> R^p) of model parameters.
> The variable y_i is the measured output for an input u_i at
time i.
> I have an algorithm to solve the above LS problem given a
nonlinear
> function
> f(u,x) and the set { (u_i,y_i), i = 1,...,N }.
> I want to use the same algorithm with minimal modifications
to solve
the
> following robust minimization problem (which Ive seen in 
a
book):
> sum_{i=1}^{N} psi(y_i - f(u_i,x)) grad_{x_k}(f) = 0,
(similar to
normal
> equations) (Eq. 2)
> k = 1, ..., p,
> where psi(x) is the Huber function:
> psi(x) = max(-k, min(x,k)),
> and grad_{x_k)(f) is the gradient vector of partial
derivaties:
df/dx(k),
> = 1,...,p.
> My problem is that I dont know how to pose the robust
problem, Eq. 2,
> (solution of the above normal equations) in a similar form
to the
least
> squares minimizations problem, Eq. 1.
> I suspect it is NOT simply
> J_r(e) = sum_{i=1}^{N} psi(y_i - f(u_i, x))^2.
===
Subject: Re: Robust Parameter Estimation
of some interest is the smoothed huber function found in
http://isi-eh.usc.es/trabajos/80_30_fullpaper.pdf
> I am trying to implement a robust estimation algorithm
instead of
nonlinear
> least squares minimization:
> My nonlinear least-squares cost if of the form:
> J(e) = sum_{i=1}^{N} [y_i - f(u_i, x)]^2, (Eq. 1)
> where f= f(u,x) is the model output for an input u and x is
the vector
(in
> R^p) of model parameters.
> The variable y_i is the measured output for an input u_i at
time i.
> I have an algorithm to solve the above LS problem given a
nonlinear
function
> f(u,x) and the set { (u_i,y_i), i = 1,...,N }.
> I want to use the same algorithm with minimal modifications
to solve the
> following robust minimization problem (which Ive seen in 
a
book):
> sum_{i=1}^{N} psi(y_i - f(u_i,x)) grad_{x_k}(f) = 0,
(similar to
normal
> equations) (Eq. 2)
> k = 1, ..., p,
> where psi(x) is the Huber function:
> psi(x) = max(-k, min(x,k)),
> and grad_{x_k)(f) is the gradient vector of partial
derivaties: df/dx(k),
k
> = 1,...,p.
> My problem is that I dont know how to pose the robust
problem, Eq. 2,
> (solution of the above normal equations) in a similar form
to the
least
> squares minimizations problem, Eq. 1.
> I suspect it is NOT simply
> J_r(e) = sum_{i=1}^{N} psi(y_i - f(u_i, x))^2.
I wish to thank all the posters ( except for Robert Kolker !)
for
their replies to my posting - I found them very enlightening.
I will
continue to monitor this thread.
>I wish to thank all the posters ( except for Robert Kolker
!) for
>their replies to my posting - I found them very
enlightening. I will
>continue to monitor this thread.
I was thinking again. If youve really been out of the info
loop, you might read the last 24 issues of _Science News_ to
get
an idea of who is doing the work you think you want to do.
/BAH
Subtract a hundred and four for e-mail.
===
Subject: Re: Jacobian elliptic function -- complex argument
(u)
Gerald> There is a sine series solution for sn(u|m) listed in
Gerald> Abramowitz & Stegun, p. 575, 16.23.1. Is the series
Gerald> legitimate if the argument u is complex?
Dont know. But you can look at the expression for
sn(x+i*y|m) to
evaluate it in terms of functions of a real arg. This is in
A&S
somewhere (dont have it handy right now).
Alternatively, you can use Landens (or 
Gauss)
transformation to
compute it. Both are given in A&S. Maxima computes the
elliptic
functions this way.
Ray
===
Subject: Re: How to get into Scientific Programming
> programming job?
Lessee here. You sprayed this message over what, five groups?
Only one of which was a jobs related news group. Well, if
your CV shows up at my company, I will be telling the HR folk
to recycle it and not even respond to you.
Socks
===
Subject: Re: How to get into Scientific Programming
> Not necessarily. Even the biggest companies seem to be
shrinking their
> research departments, and thats where youd 
probably want
to find work.
Defense industry is looking pretty good these days.
However one may have to be a citizen and have clearances for
some jobs.
The defense industry is cyclical. Its downturn was a major
part of
the early 1990s recession.
===
Subject: Re: How to get into Scientific Programming
>How did you get your degress without knowning about their
>applications of the discipline going on in the world?
>>You must not be familiar with the educational systems on
this planet.
>> I always _listened_ to the grads and instructors tell
their stories.
>> If one is curious about some discipline, there is a lot
more to
>> learn about how to do the work than in text books.
>Thats like listening to the cadre in the army. The cadre
are just like
>you, only a few months older.
> Which makes them more experienced than you. There will
always
> be people more experienced than you and there will always be
> people less experienced than you. The trick of learning is
> to tap those who have knowledge you dont have.
> The OP stated that he had masters degree in optics. In all
of
> that study time, including the undergrad work, this guy
never
> encountered a trade magazine? He never encountered
> The OP never encountered how businesses were using optics to
> make money? Computer chips, terminals, eyeglasses, contacts,
> anything thats trying to use fiber optics 
over long distance
> (telephones, LANs, any data base application), computer
games,
> windows in houses, photography, ..
> Go to a retail store and look at everything that needs some
> knowledge of optics. How could the OP set down at a
terminal,
> and type in the request without noticing all the ways that
> optics is involved in getting those ASCII bits from his
fingers
> to my terminal screen?
> The whole post smelled of troll.
> /BAH
> Subtract a hundred and four for e-mail.
I am not sure what you think I am trolling for, but I can
assure you
that my question was genuine. You also seem to think I dont
know any
applications of optics and cant find a job in 
this field. I
dont
know where you got this idea, in fact my first engineering job
that I
mentioned was as an optical engineer. I want to be doing
scientific
programming, field/industry open; this is what my question was
about.
===
Subject: Re: How to get into Scientific Programming
<snip> Go to a retail store and look at everything that needs
some
>> knowledge of optics. How could the OP set down at a
terminal,
>> and type in the request without noticing all the ways that
>> optics is involved in getting those ASCII bits from his
fingers
>> to my terminal screen?
>> The whole post smelled of troll.
>I am not sure what you think I am trolling for, but I can
assure you
>that my question was genuine.
Because, given your stated credentials, you asked a <pardon
me>
stupid question.
> ..You also seem to think I dont know any
>applications of optics and cant find a job in 
this field. I
dont
>know where you got this idea, in fact my first engineering
job that I
>mentioned was as an optical engineer. I want to be doing
scientific
>programming, field/industry open; this is what my question
was about.
Transfer to the R&D group. There is no such thing as pure
science
if youre limiting the tasks you do to programming. I
definitely
got the impression that you wanted to do programming and not
all that
other stuff (lab work) that gets you dirty.
/BAH
Subtract a hundred and four for e-mail.
===
Subject: Re: How to get into Scientific Programming
>How did you get your degress without knowning about their
>applications of the discipline going on in the world?
>>You must not be familiar with the educational systems on
this planet.
>I always _listened_ to the grads and instructors tell their
stories.
>If one is curious about some discipline, there is a lot more
to
>learn about how to do the work than in text books.
But youre assuming the grads and instructors have stories 
to
tell
about the applications of the discipline. Some do, some do
not. There
are departments whose focus is on the scientific principles
and care
not at all whether their students learn about how to do the
work.
dave
===
Subject: Re: How to get into Scientific Programming
>>How did you get your degress without knowning about their
>>applications of the discipline going on in the world?
>You must not be familiar with the educational systems on
this planet.
>>I always _listened_ to the grads and instructors tell their
stories.
>>If one is curious about some discipline, there is a lot
more to
>>learn about how to do the work than in text books.
>But youre assuming the grads and instructors have stories
to tell
>about the applications of the discipline. Some do, some do
not. There
>are departments whose focus is on the scientific principles
and care
>not at all whether their students learn about how to do the
work.
If the focus is on principles, then the training is in the
theory
area and not the application area. If the focus is on
principles
without any consideration of experimental constraints, then
the
degree is in fairy tales. If I did not want a degree in fairy
tales,
then I would fire them and find another place to 
learn.
/BAH
Subtract a hundred and four for e-mail.
===
Subject: Re: How to get into Scientific Programming
If you want to work in academia, youll eventually want to
complete
a PhD. Your prospects without the terminal degree wil be
limited.
Whatever your subject area, you should focus on computational
science, which usually involves the use of parallel or cluster
computing in the solution of larger science or engineering
problems
than can be solved on single computers running serial
programs.
Youll need to be aware of:
- standard math techniques relevant to scientific computing
(linear algebra, PDEs, ODEs, statistics, numerical methods,
etc)
- standard (serial) scientific computing techniques (FFTs,
Monte
Carlo, finite element, CFD models, etc)
- standard parallel software tools and techniques (C/C++,
fortran,
batch queuing managers, MPI, PVM, Linda, OpenMP, etc)
- standard math libraries (BLAS/LAPack/ScaLAPack/IMSL/NAG/etc)
- domain decomposition methods and tools (Metis, etc)
- performance optimization techniques and tools (compiler
tricks,
parallel debuggers, timers, tuners, etc)
- some knowledge of computer architecture (esp parallel),
operating
systems, networking, storage systems, Unix/Linux, etc.
And then youll want to be expert in some computational
domain --
physics, chemistry, applied math, or an engineering. As yet,
biology/medicine is in a world apart.
I think the prospects in this field are not bad, likely to
remain
stable, and perhaps grow in time. But it requires someone
whos
willing to put science second to computing, since youll 
spend
95% your time programming and integrating third party
software,
and little time doing science.
If you want to work in a government lab (esp. Dept of Energy),
the rules appear to be pretty much the same. However, I have
some doubts about the wisdom of working in the hydrogen bomb
industry. With the demise of the cold war, I think its
increasingly difficult to justify the expenditures necessary 
to
maintain multiple bomb labs. In the past decade, its my
impression that the labs have been shrinking and losing their
sense of direction.
Needless to say, to work for the DoE, you need to be able to
get
a US Top Secret clearance.
If you dont want to work in academia or a 
govt lab, then I
cant
tell you what your prospects are. Im not aware of a single
path
that leads to scientific programming in industry. Perhaps you
should search monster.com for matlab and see what pops up.
Randy
> Hi all,
> I was wondering if anyone could give me feedback on this
topic. I am
> 34, a US Citizen, and have a scientific background, with
Bachelors
> Degrees in Physics and Math, graduate work in Physics and
Optics
> culminating with an MS in Optics. I have worked as an
engineer, and
> then as a programmer for the last 5 years. However, it has
been
> comercial programming (using Microsoft technologies such as
Vsual C++,
> ASP, etc) and while I am thankful to have a job in the
current
> economic climate, I am also unhappy with the kind of work I
am doing.
> I would like to do some sort of scientific programming, but
I have not
> had much success applying to such jobs. They either require
an
> existing security clearance or ask for tons of prior
scientific
> programming experience.
> So my questions are:
> 1) What do you think my chances are of getting such a
scientific
> programming job? Would you expect a big difference when the
economy
> improves, hopefully in a few years?
> 2) Any suggestions of specific places where I can apply for 
a
> relatively entry level scientific programming position?
> 2) Are scientific programming jobs more secure than your
typical
> commercial position? I have heard that being over 40 and a
commercial
> programmer is a risky place to be, because of age
discrimination and
> general outsourcing. Are there many scientific programmers
who so that
> till they retire?
> If I cant find such a job, I am considering 
getting an MBA,
but I
> really think I would not enjoy that type of work nearly as
much.
> -Sam Banerjee
===
Subject: Re: How to get into Scientific Programming
<snipIf you dont want to work in academia or a 
govt lab,
then I cant
>tell you what your prospects are. Im not aware of a single
path
>that leads to scientific programming in industry. Perhaps you
>should search monster.com for matlab and see what pops up.
Wouldnt those companies who are trying to lay 
fiber optic
cable across oceans use programming to develop their products?
What about hardware design? Has there been any work done using
fiber optics? Comm (phones and cable) would 
benefit from fiber
applications because of bandwidth and leakage.
/BAH
<pins>
/BAH
Subtract a hundred and four for e-mail.
===
Subject: Re: How to get into Scientific Programming
> <snipIf you dont want to work in academia or a 
govt lab,
then I cant
>tell you what your prospects are. Im not aware of a single
path
>that leads to scientific programming in industry. Perhaps you
>should search monster.com for matlab and see what pops up.
> Wouldnt those companies who are trying to lay 
fiber optic
> cable across oceans use programming to develop their
products?
> What about hardware design? Has there been any work done
using
> fiber optics? Comm (phones and cable) would 
benefit from fiber
> applications because of bandwidth and leakage.
They do it using commercial packages. People are right - the
market
for scientific programming is extremely narrow. So, your
choice is
either apply to one of existing companies like MathWork,
Fluent or
MolecularSimulation or try to push your own package.
Old Pif
===
Subject: Re: How to get into Scientific Programming
>> <snip>If you dont want to work in academia or a 
govt
lab, then I cant
>>tell you what your prospects are. Im not aware of a 
single
path
>>that leads to scientific programming in industry. Perhaps 
you
>>should search monster.com for matlab and see what pops up.
>> Wouldnt those companies who are trying to lay 
fiber optic
>> cable across oceans use programming to develop their
products?
>> What about hardware design? Has there been any work done
using
>> fiber optics? Comm (phones and cable) would 
benefit from fiber
>> applications because of bandwidth and leakage.
>They do it using commercial packages.
And who makes those commercial packages? Who takes the tools
offered in a package and knows how to use them to do the
work that is specific to a particular discipline. If those
packages are written for only one ßavor of computing, then
somebodies know enough about the science to provide the
tools.
> ..People are right - the market
>for scientific programming is extremely narrow. So, your
choice is
>either apply to one of existing companies like MathWork,
Fluent or
>MolecularSimulation or try to push your own package.
But it takes knowledge of the science to make those tools do
anything useful. Isnt this manipulation of the tools called
scientific programming? The people who write the code that
produce those tools arent doing any scientific 
programming;
theyre doing software development that manages to convince
a computer to add more correctly.
/BAH
Subtract a hundred and four for e-mail.
===
Subject: Re: How to get into Scientific Programming
>> <snip>
>>If you dont want to work in academia or a 
govt lab, then
I cant
>>tell you what your prospects are. Im not aware of a 
single
path
>>that leads to scientific programming in industry. Perhaps 
you
>>should search monster.com for matlab and see what pops up.
>>
>> Wouldnt those companies who are trying to lay 
fiber optic
>> cable across oceans use programming to develop their
products?
>> What about hardware design? Has there been any work done
using
>> fiber optics? Comm (phones and cable) would 
benefit from fiber
>> applications because of bandwidth and leakage.
>>
>They do it using commercial packages.
> And who makes those commercial packages?
Scientific programmers do. However, the total market share of
these
jobs is miniscule compare to the rest.
> Who takes the tools
> offered in a package and knows how to use them to do the
> work that is specific to a particular discipline. If those
> packages are written for only one ßavor of computing, then
> somebodies know enough about the science to provide the
> tools.
Those packages have been created with the sole purpose to
relieve
engineers and scientists from programming. And the goal has
been
attained.
> ..People are right - the market
>for scientific programming is extremely narrow. So, your
choice is
>either apply to one of existing companies like MathWork,
Fluent or
>MolecularSimulation or try to push your own package.
> But it takes knowledge of the science to make those tools do
> anything useful. Isnt this manipulation of the tools 
called
> scientific programming?
No, it is not. Those people are users, not programmers. They
are
specialists in their fields and often dont know 
how to write
a line
of code.
> The people who write the code that
> produce those tools arent doing any 
scientific programming;
> theyre doing software development that manages to 
convince
> a computer to add more correctly.
See above.
Old Pif
===
Subject: Re: How to get into Scientific Programming
>> But it takes knowledge of the science to make those tools
do
>> anything useful. Isnt this manipulation of the tools
called
>> scientific programming?
>No, it is not. Those people are users, not programmers. They
are
>specialists in their fields and often dont 
know how to write
a line
>of code.
But what if youre using a commercial package like Igor Pro,
and within
that environment you write many inter-connected functions
that involve
variables, loops, conditionals, and elements of graphical
interfaces, for
the analysis and display of data?
--
Let us learn to dream, gentlemen, then perhaps we shall find
the
truth... But let us beware of publishing our dreams before
they have been
put to the proof by the waking understanding. -- Friedrich
August
Kekul.8e
===
Subject: Re: How to get into Scientific Programming
>> But it takes knowledge of the science to make those tools
do
>> anything useful. Isnt this manipulation of the tools
called
>> scientific programming?
>No, it is not. Those people are users, not programmers. They
are
>specialists in their fields and often dont 
know how to write
a line
>of code.
> But what if youre using a commercial package like Igor
Pro, and within
> that environment you write many inter-connected functions
that involve
> variables, loops, conditionals, and elements of graphical
interfaces, for
> the analysis and display of data?
Well, the major issue is whether employers see such skills as
a
profession. The typical combination is as somebody above
explained a
field and programming on the top. Say, CFD and knowledge of
some
packages, which includes skills to write customizations. If
you apply
for a position in computational ßuid dynamics, saying in your
resume
that you did scientific programming, which include everything
that you
mentioned, in say CHARM framework, the chance to get an
interview is
nil.
The reason is that in computationally intensive fields like
computational chemistry of physics the fact that you are able
to
program is something close to the statement that you can write
English. Everybody can do that. Some do that better some
worse but by
and large it is a common knowledge. It does not give you any
special
advantage.
There is very narrow niche that is filled by companies that
produce
such packages and this is very interesting work. But the
amount of
people employed is very small. University guys who spin off
small
companies that in time sustained the market stress have
started most
of them in the past. Usually they grow around the initial
core people
that are pretty stable in time. So, you have to be in the
right time
in the right place.
Old Pif
===
Subject: Re: How to get into Scientific Programming
> But what if youre using a commercial package like Igor
Pro, and within
> that environment you write many inter-connected functions
that involve
> variables, loops, conditionals, and elements of graphical
interfaces, for
> the analysis and display of data?
You mean, like, fiddling with a hundred Visual Basic lines?
That is not programming, my man. :p
But hey, with what ifs, we could build another America,
right?
===
Subject: Re: How to get into Scientific Programming
><snip>If you dont want to work in academia or a 
govt lab,
then I cant
>>tell you what your prospects are. Im not aware of a 
single
path
>>that leads to scientific programming in industry. Perhaps 
you
>>should search monster.com for matlab and see what pops up.
>Wouldnt those companies who are trying to lay 
fiber optic
>cable across oceans use programming to develop their
products?
>What about hardware design? Has there been any work done
using
>fiber optics? Comm (phones and cable) would 
benefit from fiber
>applications because of bandwidth and leakage.
One thing thats become possible now with modern computers 
is
to design an
alloy from first principles, so you can get the properties you
want on the
first melt. Thats been done at least once so 
far, and will
surely be
done again and with increasing frequency.
Meteorology always has computer models.
Lots of engineering applications -- aerodynamics of things,
for instance.
--
Let us learn to dream, gentlemen, then perhaps we shall find
the
truth... But let us beware of publishing our dreams before
they have been
put to the proof by the waking understanding. -- Friedrich
August
Kekul.8e
===
Subject: Re: How to get into Scientific Programming
> One thing thats become possible now with modern computers
is to design an
> alloy from first principles, so you can get the properties
you want on the
> first melt. Thats been done at least once so 
far, and will
surely be
> done again and with increasing frequency.
> Meteorology always has computer models.
> Lots of engineering applications -- aerodynamics of things,
for instance.
If you meant neural nets and genetic algorithms, one can add
forecasting of financial markets. Anyway, here are some
remarks as to
where one can try working on such-like things. Suppose one
has a
choice between:
1. Finding the most efficient airfoil profile 
using, say,
genetic
algorithms, with annealing, and working on it with a Prof at
a Uni
2. Doing more or less the same, in a large aircraft
corporation,
working under a technocrat
I would prefer choice #1. One might earn a bit more by
choosing #2,
but one can expect to learn more professionally interesting
things
from a Prof, rather than from a technocrat. In addition, your
achievements at the edu will be public domain, rather than
classified
stuff, known to a few, at the com place. Finally, talking
about
customer satisfaction (a-la Dilbert speak), I think that
satisfying
taxpayers, that includes yourself, is a bit more appealing
than
satisfying some stock holders in an aircraft factory.
--Baruch
===
Subject: Re: How to get into Scientific Programming
[...]
> I would prefer choice #1. One might earn a bit more by
choosing #2,
> but one can expect to learn more professionally interesting
things
> from a Prof, rather than from a technocrat. In addition,
your
> achievements at the edu will be public domain, rather than
classified
> stuff, known to a few, at the com place.
Probably wont be public domain - most universities take
possession
of intellectual property and market it. In addition, many
research
programs at universities are conducted under contract
(government
or commercial) and, while the research results might be
shared to
a greater or lesser extent, the end result is owned by
someone.
On the other hand, many companies publish their research and,
of
course, patents have to reveal everything to the world.
> Finally, talking about
> customer satisfaction (a-la Dilbert speak), I think that
satisfying
> taxpayers, that includes yourself, is a bit more appealing
than
> satisfying some stock holders in an aircraft factory.
No real difference between tax payers and stock holders - in
both
cases, they pay the bills and deserve something. -Wm
===
Subject: Re: How to get into Scientific Programming
> [...]
> I would prefer choice #1. One might earn a bit more by
choosing #2,
> but one can expect to learn more professionally interesting
things
> from a Prof, rather than from a technocrat. In addition,
your
> achievements at the edu will be public domain, rather than
classified
> stuff, known to a few, at the com place.
> Probably wont be public domain - most universities take
possession
> of intellectual property and market it. In addition, many
research
> programs at universities are conducted under contract
(government
> or commercial) and, while the research results might be
shared to
> a greater or lesser extent, the end result is owned by
someone.
The state is an owner too. It represents taxpayers in this
respect.
Some (non necessarily, many) R&D programs in Universities are
conducted under commercial contracts. Still, in my parts of
the world,
Universities still have a major degree of academic freedom,
supported
by state funds. Governments do recognize the importance of
long term
investment that the private sector wouldnt even consider if
it cant
see immediate profits there. Think about the human genome
project.
AFAIK, it is an Academic/Governmental initiative. Hope
Universities
will not turn into commercial contractors. That will end their
academic freedom.
> On the other hand, many companies publish their research
and, of
> course, patents have to reveal everything to the world.
Oh great! Imagine, F = m*a, in a patent. It will be $5 each
and every
Einstein lived before our commodization era.
> Finally, talking about
> customer satisfaction (a-la Dilbert speak), I think that
satisfying
> taxpayers, that includes yourself, is a bit more appealing
than
> satisfying some stock holders in an aircraft factory.
> No real difference between tax payers and stock holders -
in both
> cases, they pay the bills and deserve something. -Wm
Some difference: when people buy shares in a factory they
dont mind
if working conditions there are slave-like. They buy shares
for
profit, selling them when there is no profit. Not 
so, with
taxpayers.
They dont expect profits, they expect to 
improve quality of
life.
Well, stock holders are taxpayers as well. However, their
profit
seeking might come at the account of somebody elses quality
of life.
Not sure if the inventor of Penicillin profited much from it.
A lot of
taxpayers certainly did. Now, back to the present. Expensive
medicines
are more profitable. An interesting factor, 
isnt it?
--Baruch
===
Subject: Re: How to get into Scientific Programming
> No real difference between tax payers and stock holders -
in both
> cases, they pay the bills and deserve something. -Wm
stock holders dont pay the bills. They pay the person they
bought the
stock from.
===
Subject: Re: How to get into Scientific Programming
>
>
> No real difference between tax payers and stock holders -
in both
> cases, they pay the bills and deserve something. -Wm
>
>
>
> stock holders dont pay the bills. They pay the person 
they
bought the
> stock from.
Just where do you think the revenue from an IPO goes, and how
it is used?
Harry C.
===
Subject: Re: How to get into Scientific Programming
omission9 replied:
>> No real difference between tax payers and stock holders -
in both
>> cases, they pay the bills and deserve something. -Wm
> stock holders dont pay the bills. They pay the person 
they
bought the
> stock from.
I think that lots of investors in dot.coms might argue this
with you.
Rich
===
Subject: Re: How to get into Scientific Programming
>> One thing thats become possible now with modern 
computers
is to design
an
>> alloy from first principles, so you can get the properties
you want on
the
>> first melt. Thats been done at least once so 
far, and will
surely be
>> done again and with increasing frequency.
>> Meteorology always has computer models.
>> Lots of engineering applications -- aerodynamics of
things, for
instance.
>If you meant neural nets and genetic algorithms, one can add
>forecasting of financial markets. Anyway, here are some
remarks as to
>where one can try working on such-like things. Suppose one
has a
>choice between:
>1. Finding the most efficient airfoil profile 
using, say,
genetic
>algorithms, with annealing, and working on it with a Prof at
a Uni
>2. Doing more or less the same, in a large aircraft
corporation,
>working under a technocrat
>I would prefer choice #1. One might earn a bit more by
choosing #2,
>but one can expect to learn more professionally interesting
things
>from a Prof, rather than from a technocrat. In addition, your
>achievements at the edu will be public domain, rather than
classified
>stuff, known to a few, at the com place. Finally, talking
about
>customer satisfaction (a-la Dilbert speak), I think that
satisfying
>taxpayers, that includes yourself, is a bit more appealing
than
>satisfying some stock holders in an aircraft factory.
I see youre using a computer which runs software to make
this post.
Arent you glad that we didnt have the same 
attitude that
you do?
/BAH
Subtract a hundred and four for e-mail.
===
Subject: Re: How to get into Scientific Programming
>>where one can try working on such-like things. Suppose one
has a
>>choice between:
>>1. Finding the most efficient airfoil profile 
using, say,
genetic
>>algorithms, with annealing, and working on it with a Prof
at a Uni
>>2. Doing more or less the same, in a large aircraft
corporation,
>>working under a technocrat
>>I would prefer choice #1. One might earn a bit more by
choosing #2,
>>but one can expect to learn more professionally interesting
things
>>from a Prof, rather than from a technocrat. In addition,
your
If by professionally interesting things you mean talking about
aircraft, rather than making them. (In point of fact, more
likely
talking about some theoretical stuff that is not used in
practice.)
>>achievements at the edu will be public domain, rather than
classified
>>stuff, known to a few, at the com place. Finally, talking
about
Patents go into the public domain eventually.
>>customer satisfaction (a-la Dilbert speak), I think that
satisfying
>>taxpayers, that includes yourself, is a bit more appealing
than
>>satisfying some stock holders in an aircraft factory.
And satisfying people who actually use or ßy aircraft. It all
seems a
clear win for joining an aircraft corporation if aircraft are
your
interest. Universities dont have any real input into 
aircraft
creation. For some fields the decision may be less obvious.
In this context, the Encyclopaedia Britannica helpfully
points out that
the Wright brothers were the only members of the Wright
family who did
not attend college!
(Of course they did use certain university resources in some
of their
research.)
- Gerry Quinn
- Gerry Quinn
===
Subject: Re: How to get into Scientific Programming
>If you meant neural nets and genetic algorithms, one can add
>forecasting of financial markets. Anyway, here are some
remarks as to
>where one can try working on such-like things. Suppose one
has a
>choice between:
>1. Finding the most efficient airfoil profile 
using, say,
genetic
>algorithms, with annealing, and working on it with a Prof at
a Uni
>2. Doing more or less the same, in a large aircraft
corporation,
>working under a technocrat
>I would prefer choice #1. One might earn a bit more by
choosing #2,
>but one can expect to learn more professionally interesting
things
>from a Prof, rather than from a technocrat. In addition, your
>achievements at the edu will be public domain, rather than
classified
>stuff, known to a few, at the com place. Finally, talking
about
>customer satisfaction (a-la Dilbert speak), I think that
satisfying
>taxpayers, that includes yourself, is a bit more appealing
than
>satisfying some stock holders in an aircraft factory.
> I see youre using a computer which runs software to make
this post.
> Arent you glad that we didnt have the same 
attitude that
you do?
> /BAH
I remember using Bitnet in early 80s, at a Uni. Not sure the
software
there was all commercial. AFAIK, networks started as a gov
initiative,
not a com initiative. Arent you happy that those gifted
people in gov
labs (including, of course the gifted WWW people of CERN, of
course)
had invented the internet for us? Well they didnt care much
about
on-line shopping etc. That can be taken by the com people
provided
they take care about security (like using the public key
encription
that was developed at MIT...).
--Baruch
===
Subject: Re: How to get into Scientific Programming
>>If you meant neural nets and genetic algorithms, one can add
>>forecasting of financial markets. Anyway, here are some
remarks as to
>>where one can try working on such-like things. Suppose one
has a
>>choice between:
>>1. Finding the most efficient airfoil profile 
using, say,
genetic
>>algorithms, with annealing, and working on it with a Prof
at a Uni
>>2. Doing more or less the same, in a large aircraft
corporation,
>>working under a technocrat
>>I would prefer choice #1. One might earn a bit more by
choosing #2,
>>but one can expect to learn more professionally interesting
things
>>from a Prof, rather than from a technocrat. In addition,
your
>>achievements at the edu will be public domain, rather than
classified
>>stuff, known to a few, at the com place. Finally, talking
about
>>customer satisfaction (a-la Dilbert speak), I think that
satisfying
>>taxpayers, that includes yourself, is a bit more appealing
than
>>satisfying some stock holders in an aircraft factory.
>> I see youre using a computer which runs software to make
this post.
>> Arent you glad that we didnt have the 
same attitude that
you do?
>> /BAH
>I remember using Bitnet in early 80s, at a Uni. Not sure the
software
>there was all commercial. AFAIK, networks started as a gov
initiative,
>not a com initiative. Arent you happy that those gifted
people in gov
>labs (including, of course the gifted WWW people of CERN, of
course)
>had invented the internet for us? Well they didnt care 
much
about
>on-line shopping etc. That can be taken by the com people
provided
>they take care about security (like using the public key
encription
>that was developed at MIT...).
Honey, if you had been working at ORNL in 1970, you might
have had access to the network that TOPS-10 and some
others built for them. DEC was a publicly traded company.
And you want to spout a 1980 date at me????? Kids!
/BAH
Subtract a hundred and four for e-mail.
===
Subject: Re: How to get into Scientific Programming
>I remember using Bitnet in early 80s, at a Uni. Not sure the
software
>there was all commercial. AFAIK, networks started as a gov
initiative,
>not a com initiative. Arent you happy that those gifted
people in gov
>labs (including, of course the gifted WWW people of CERN, of
course)
>had invented the internet for us? Well they didnt care 
much
about
>on-line shopping etc. That can be taken by the com people
provided
>they take care about security (like using the public key
encription
>that was developed at MIT...).
> Honey, if you had been working at ORNL in 1970, you might
> have had access to the network that TOPS-10 and some
> others built for them. DEC was a publicly traded company.
> And you want to spout a 1980 date at me????? Kids!
> /BAH
(from: http://www.dei.isep.ipp.pt/docs/arpa--4.html)
The environment we were operating in was one of open
research. The
only payoff available was to have good work recognized and
used.
Software was generally considered free. Openness wasnt an
option; it
just was. (Crocker, 1993c)
It is about Arpanet. Sure, there must have been a lot of
contractors
involved. But who had the seminal idea, so to speak.
BTW, no witty remarks about WWW-CERN and MIT, I mentioned?
--Baruch
===
Subject: Re: Simplifying logical proposition
>
> I would like to know if anyone could help me understanding
how Maple
did the
> following simplification (using bsimp):
Proposition: (not p or q) and ((q and not r) or (r and not q))
> Simplified proposition: r and not p and not q
>
> My Maple 9 bsimp gives
> (r and not p and not q) or (q and not r)
>
> I dont think bsimp has been changed in years. What Maple
are you
using?
> Something must have changed since you both got different
and *wrong*
> results. Perhaps a few other Maple users can pitch in to
establish
> whether Maple (and, indirectly, Matlab) needs to be
recalled!?
> btw, the correct result (via paper & pencil method) is (for
x_ = not x),
> (q_ and r) or (q and r_)
Thats correct in the sense of incorrect. Suppose we
have p, not-q, r. Then the original formula is
(not p or q) and ((q and not r) or (r and not q))
= (false or false) and ((false and false) or (true and true))
= false
Carls is
(r and not p and not q) or (q and not r)
= (true and false and true) or (false and false)
= false
And yours is
(not q and r) or (q and not r)
= (true and true) or (false and false)
= true
So: on the basis of an (obviously, I may add) incorrect
calculation, you suggest that Maple needs to be recalled
(as if that would be the right response if it *were*
broken!). Anyone would have thought you work for the
competition, or something.
--
Gareth McCaughan
.sig under construc
===
Subject: Re: Simplifying logical proposition
<Pine.GSO.4.33.0310132053530.12493-100000@naxos.math.udel.edu>
<3F8DD8AC.53E9A542@Xsdynamix.com Perhaps a few other Maple
users can pitch in to establish
> whether Maple (and, indirectly, Matlab) needs to be
recalled!?
> Dr.B.Voh
> Applied Algorithms http://sdynamix.com
Since you work for one of the competitors, your comment above
must be
taken with a grain of salt.
===
Subject: can someone help me with this problem?
hi, can someone please help me with this problem?
If A is a finite set and f:A-->A is 1-1,then f is onto (use
the fact
that a proper subset of a finite set has smaller size)
===
Subject: Need help for this integral
by support1.mathforum.org (8.11.6/8.11.6/The Math Forum,
$Revision:
1.9 primary) id h9KG4vk29787;
Id like to calculate the value of the integral
int_{-infty} ^infty exp(ikx) Ln(1/Abs(x-x)) dx.
Is it possible to calculate the integral analytically?
Can anyone give me hints for how to solve such case?
David
===
Subject: Greens function in spectral domain
by support1.mathforum.org (8.11.6/8.11.6/The Math Forum,
$Revision:
1.9 primary) id h9KJPoM11758;
I would like to find the 2D Greens function in 
spectral
domain. Does
anyone has the answer ablout it? or know where I can find it?
David