gsl optimization routines

gsl optimization routines

[Dan, Ho-Jin]

Hi there,
I installed the gsl library and found that there are missing features of 
the optimization, i.e.,
It can't handle the constraints. So I carefully consider filling the 
missing routines such as
SQP, the method of the feasible directions etc. If there are not on 
going project on optimization
I will start hack the code. But I can't guarantee the efficieny and 
robustness of the algoritm at the first time.
moreover, my english is poor so that the manual from the work must be 
reviewed by someone else.
Is there any suggestion ?

Re: gsl optimization routines

[Brian Gough]

Dan, Ho-Jin writes:
 > Hi there, I installed the gsl library and found that there are
 > missing features of the optimization, i.e., It can't handle the
 > constraints. So I carefully consider filling the missing routines
 > such as SQP, the method of the feasible directions etc. If there
 > are not on going project on optimization I will start hack the
 > code. But I can't guarantee the efficieny and robustness of the
 > algoritm at the first time.  moreover, my english is poor so that
 > the manual from the work must be reviewed by someone else.  Is
 > there any suggestion ?

Hi,

Thanks for the messsage.  I'm not aware of anyone else working on
constrained optimisation so you're welcome to work on that.  We can
sort out any documentation problems.

An important step is to see if there is an interface which can support
a large class of algorithms as in the multifit/ and multmin/
directories.  I wrote some notes in the document doc/gsl-design.texi
which might be helpful there.

best regards,

Brian Gough

Re: gsl optimization routines

[Dan, Ho-Jin]

Okay, I'll figure out the gsl stucture, check reusable functions and survey
recently published papers. First of all, nonlinear programming(NP) will
be focused on with a dense storage(small to medium size problems), and,
if needed, linear programming(LP) will be implemented. actually, I am
not interested in LP since the research area of mine is the structural 
optimization.

sincerely yours

Brian Gough wrote:

>Dan, Ho-Jin writes:
> > Hi there, I installed the gsl library and found that there are
> > missing features of the optimization, i.e., It can't handle the
>An important step is to see if there is an interface which can support
>a large class of algorithms as in the multifit/ and multmin/
>directories.  I wrote some notes in the document doc/gsl-design.texi
>which might be helpful there.
>
>best regards,
>
>Brian Gough
>
>

Re: gsl optimization routines

[Brian Gough]

Dan, Ho-Jin writes:
 > Okay, I'll figure out the gsl stucture, check reusable functions and survey
 > recently published papers. First of all, nonlinear programming(NP) will
 > be focused on with a dense storage(small to medium size problems), and,
 > if needed, linear programming(LP) will be implemented. actually, I am
 > not interested in LP since the research area of mine is the structural 
 > optimization.

Yes, routines for nonlinear programming are needed.  For linear
programming there is an excellent GNU package already called GLPK (GNU
Linear Programming Kit) written by Andrew Makhorin which people can
use.

Brian

Re: gsl optimization routines

[Brian Gough]

Dan, Ho-Jin writes:
 > Hi there, I installed the gsl library and found that there are
 > missing features of the optimization, i.e., It can't handle the
 > constraints. So I carefully consider filling the missing routines
 > such as SQP, the method of the feasible directions etc. If there
 > are not on going project on optimization I will start hack the
 > code. But I can't guarantee the efficieny and robustness of the
 > algoritm at the first time.  moreover, my english is poor so that
 > the manual from the work must be reviewed by someone else.  Is
 > there any suggestion ?

Hi,

Thanks for the messsage.  I'm not aware of anyone else working on
constrained optimisation so you're welcome to work on that.  We can
sort out any documentation problems.

An important step is to see if there is an interface which can support
a large class of algorithms as in the multifit/ and multmin/
directories.  I wrote some notes in the document doc/gsl-design.texi
which might be helpful there.

best regards,

Brian Gough

gsl optimization routines

[Dan, Ho-Jin]

Hi there,
I installed the gsl library and found that there are missing features of 
the optimization, i.e.,
It can't handle the constraints. So I carefully consider filling the 
missing routines such as
SQP, the method of the feasible directions etc. If there are not on 
going project on optimization
I will start hack the code. But I can't guarantee the efficieny and 
robustness of the algoritm at the first time.
moreover, my english is poor so that the manual from the work must be 
reviewed by someone else.
Is there any suggestion ?

Re: gsl optimization routines

[Brian Gough]

Dan, Ho-Jin writes:
 > Okay, I'll figure out the gsl stucture, check reusable functions and survey
 > recently published papers. First of all, nonlinear programming(NP) will
 > be focused on with a dense storage(small to medium size problems), and,
 > if needed, linear programming(LP) will be implemented. actually, I am
 > not interested in LP since the research area of mine is the structural 
 > optimization.

Yes, routines for nonlinear programming are needed.  For linear
programming there is an excellent GNU package already called GLPK (GNU
Linear Programming Kit) written by Andrew Makhorin which people can
use.

Brian

Re: gsl optimization routines

[Dan, Ho-Jin]

Okay, I'll figure out the gsl stucture, check reusable functions and survey
recently published papers. First of all, nonlinear programming(NP) will
be focused on with a dense storage(small to medium size problems), and,
if needed, linear programming(LP) will be implemented. actually, I am
not interested in LP since the research area of mine is the structural 
optimization.

sincerely yours

Brian Gough wrote:

>Dan, Ho-Jin writes:
> > Hi there, I installed the gsl library and found that there are
> > missing features of the optimization, i.e., It can't handle the
>An important step is to see if there is an interface which can support
>a large class of algorithms as in the multifit/ and multmin/
>directories.  I wrote some notes in the document doc/gsl-design.texi
>which might be helpful there.
>
>best regards,
>
>Brian Gough
>
>