Report on Numerical Analysis Techniques Session [message #60630] |
Thu, 29 May 2008 13:54 |
David Fanning
Messages: 11724 Registered: August 2001
|
Senior Member |
|
|
Folks,
The good folks at ITTVIS have been giving a series of occasional
seminars for IDL users in Boulder (and perhaps elsewhere, I don't
know). But, at least in Boulder, these are extremely well attended,
and not just because they were giving out free T-shirts at the end.
There has been some fairly good information coming out of them.
I attended the "Numerical Analysis Techniques in IDL" session today
and thought I would give you a report, just to keep everyone up to
date.
The purpose of the seminar was to "learn how to get the most out
of your data using IDL�s built-in math and statistics libraries."
Then there was some information on how to unlock the IDL Analyst
product. This caused some people in my group to feel like this
would be an extended infomercial for IDL Analyst, but as it turned
out, IDL Analyst was barely mentioned and all the examples used
normal IDL routines.
Mark Piper gave most of the presentation. In my travels in the IDL
community I have heard nothing but great reports of Mark's teaching
abilities, and now I know why. He is friendly, knowledgeable, an
expert communicator, and he can keep a smile on his face even when
his first unrehearsed foray into an iTool modification completely
crashed IDL. (Oh, well, I've seen worse with the iContour command.)
One of my gripes with IDL examples is that they are often
so contrived as to me meaningless or misleading when it comes
to applying the example to real-world data. None of that here.
Mark used real data, simple enough to use in this setting, but
sophisticated enough to make you think you could probably try
this at home.
The problem with teaching data analysis is the fear you encounter
when you face a room full of people who's livelihood is nothing
*but* data analysis, and you have to wave your hands at the little
bit you *think* you know. (I speak of myself here, not Mark.)
Mark is more knowledgeable about this stuff than I am, but clearly
there were people in the audience who wrote their graduate theses
on this stuff. Mark handled it all with grace and good humor, and
with a minimum amount of handwaving.
When Mark finished, a member of the IDL Professional Development
group got up to do a short presentation on "particle swarm
optimisation". I didn't follow too many of the technical details,
but basically your throw a bizillion random "particles" at a
curve fitting or minimization problem, and they can coalesce into
a solution after some finite number of iterations. The examples
were neat, and the code was sparse. (IDL always amazes me!)
The most interesting part of this presentation to me personally
was that this person had apparently never heard of the IDL
Workbench (or perhaps he had) and had never used it. He showed
us the code using the VI editor. Go figure. This part of the
presentation seemed rather free of marketing input. :-)
The final presentation of the day was by Peter Messmer of
Tech-X Corporation. His company has developed some IDL code
that allows you to take advantage of the processing capability
built into the 128 processors of your graphics card! Wow.
For the right application, this could be extremely useful.
Best of all, Tech-X is making it available free (using a GPL
license) for people to take advantage of. They plan to make
their money by using this as an introduction to the companies
other parallel processing products.
http://www.txcorp.com/technologies/GPULib/index.php
Overall, I thought it was a useful seminar. I learned a couple
of things, had my prejudices about iTools confirmed, and spent
a pleasant morning in good company. I appreciate ITTVIS doing
this kind of outreach. If you have a chance to attend a seminar,
I would highly recommend it.
Cheers,
David
--
David Fanning, Ph.D.
Fanning Software Consulting, Inc.
Coyote's Guide to IDL Programming (www.dfanning.com)
Sepore ma de ni thui. ("Perhaps thou speakest truth.")
|
|
|