| Genetic algorithms and curve fitting [message #34521] |
Thu, 27 March 2003 13:26  |
robert.dimeo
Messages: 42
Registered: November 2001
|
Member |
|
|
Hi,
Over the past few years I have been writing IDL programs to analyze
data by fitting to models using standard least-squares methods. The
problems that we face here, however, sometimes are not amenable to the
least-squares approach and, having multi-modal chi-squared surfaces,
could benefit from a more robust global optimization. I have been
working on a hybrid genetic algorithm/least squares function that
appears to work "pretty well" for model fitting.
Has anyone on this newsgroup you used GAs or GA hybrids for data
fitting? I am seeking someone with whom I can "compare notes" and
discuss what works well and what does not. I am happy to share this
code with anyone (especially if you're interested in helping me
improve it!)
Many thanks in advance!
Rob
----------------------
Robert M. Dimeo, Ph.D.
NIST Center for Neutron Research
National Institute of Standards and Technology
100 Bureau Drive - Stop 8562 Gaithersburg, MD 20899-8562
Telephone: (301) 975-8135 FAX: (301) 921-9847
www.ncnr.nist.gov/staff/dimeo
|
|
|
|
| Re: Genetic algorithms and curve fitting [message #34608 is a reply to message #34521] |
Fri, 28 March 2003 08:10  |
Allard de Wit
Messages: 7
Registered: December 2002
|
Junior Member |
|
|
Rob Dimeo wrote:
> Hi,
>
> Over the past few years I have been writing IDL programs to analyze
> data by fitting to models using standard least-squares methods. The
> problems that we face here, however, sometimes are not amenable to the
> least-squares approach and, having multi-modal chi-squared surfaces,
> could benefit from a more robust global optimization. I have been
> working on a hybrid genetic algorithm/least squares function that
> appears to work "pretty well" for model fitting.
>
> Has anyone on this newsgroup you used GAs or GA hybrids for data
> fitting? I am seeking someone with whom I can "compare notes" and
> discuss what works well and what does not. I am happy to share this
> code with anyone (especially if you're interested in helping me
> improve it!)
>
> Many thanks in advance!
>
> Rob
> ----------------------
> Robert M. Dimeo, Ph.D.
> NIST Center for Neutron Research
> National Institute of Standards and Technology
> 100 Bureau Drive - Stop 8562 Gaithersburg, MD 20899-8562
> Telephone: (301) 975-8135 FAX: (301) 921-9847
>
> www.ncnr.nist.gov/staff/dimeo
Hi Rob,
I do not have an extensive knowledge of GA techniques but for the
relatively simple optimisation problems that I have used GA's for
I used a so-called micro-GA which worked pretty well. I implemented
the micro-GA in IDL code, you can find the code at
www.xs4all.nl/~ajwwag/
As it this GA was my first attempt to write something in IDL using
object-oriented programming, the programming style may be somewhat
clumsy, any comments on the code (errors/fixes) are welcome.
Maybe, it is usefull for you.
best regards,
Allard de Wit
|
|
|
|
| Re: Genetic algorithms and curve fitting [message #34843 is a reply to message #34608] |
Wed, 23 April 2003 11:57 |
pford
Messages: 33
Registered: September 1996
|
Member |
|
|
"A. De Wit" <adewit@operamail.com> wrote in message news:<3e8473ac$0$49109$e4fe514c@news.xs4all.nl>...
> Rob Dimeo wrote:
>> Hi,
>>
>> Over the past few years I have been writing IDL programs to analyze
>> data by fitting to models using standard least-squares methods. The
>> problems that we face here, however, sometimes are not amenable to the
>> least-squares approach and, having multi-modal chi-squared surfaces,
>> could benefit from a more robust global optimization. I have been
>> working on a hybrid genetic algorithm/least squares function that
>> appears to work "pretty well" for model fitting.
>>
>> Has anyone on this newsgroup you used GAs or GA hybrids for data
>> fitting? I am seeking someone with whom I can "compare notes" and
>> discuss what works well and what does not. I am happy to share this
>> code with anyone (especially if you're interested in helping me
>> improve it!)
>>
>> Many thanks in advance!
>>
>> Rob
>> ----------------------
>> Robert M. Dimeo, Ph.D.
>> NIST Center for Neutron Research
>> National Institute of Standards and Technology
>> 100 Bureau Drive - Stop 8562 Gaithersburg, MD 20899-8562
>> Telephone: (301) 975-8135 FAX: (301) 921-9847
>>
>> www.ncnr.nist.gov/staff/dimeo
>
> Hi Rob,
>
> I do not have an extensive knowledge of GA techniques but for the
> relatively simple optimisation problems that I have used GA's for
> I used a so-called micro-GA which worked pretty well. I implemented
> the micro-GA in IDL code, you can find the code at
> www.xs4all.nl/~ajwwag/
>
> As it this GA was my first attempt to write something in IDL using
> object-oriented programming, the programming style may be somewhat
> clumsy, any comments on the code (errors/fixes) are welcome.
> Maybe, it is usefull for you.
>
> best regards,
>
> Allard de Wit
I used a simple GA in the past written in C and called using
CALL_EXTERNAL. I will be happy to send this to you. I am in the SLOW
process of converting it to DLM on MacOS X. I have to relearn IDL and
learn Unix and gcc, etc.
Patrick Ford, MD
Baylor College of Medicine
pford@*nospam*bcm.tmc.edu
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
