[Bonmin] bonmin and ampl external function

Pierre Bonami pierre.bonami at gmail.com
Sun Mar 4 10:34:24 EST 2007


Dear Roy,
I just tried to verify and I can use imported ampl functions with  
bonmin on my machine without problem.
Please submit a ticket with as precise a description as you can do so  
that we can try to reproduce it. Please also include the model file  
so that you are using.
Best regards,
Pierre
On Mar 3, 2007, at 7:56 PM, Roy Chen wrote:

> Dear Andreas
> I tried function ginv() with bonmin, which is an example in  
> funcadd.c. The followings are two experiments I did:
> Experiment 1:
> ampl: function ginv(Reals);
> ampl: display {i in -3...3} ginv(i);
> ginv(i) [*] :=
> -3  -0.333333
> -2  -0.5
> -1  -1
>  0   0
> ;
> Experiment 2:
> ampl: reset;
> ampl: function ginv(Reals);
> ampl: var x;
> ampl: minimize Z:x*x+x+1+ginv(x);
> ampl: subject to cons:x>=0.8;
> ampl: option solver bonmin;
> ampl: solve;
> bonmin:
> Cannot load library /home/chen11q/AMPL_Learning/ 
> ampl_external_function/ASL/solvers/funclink/amplfunc.dll:
> /home/chen11q/AMPL_Learning/ampl_external_function/ASL/solvers/ 
> funclink/amplfunc.dll: cannot open shared object file: No such file  
> or directory
> function ginv not available
> user-defined function not available
> exit code 255
> <BREAK>
> What I do not quite understand is: as you said in this example, the  
> first and second derivatives of the function ginv() has been  
> defined properly in C code, then why couldn't gin()still be used  
> with Bonmin in AMPL? I am just trying to run this small example to  
> make sure that once I can successfully write the first and second  
> derivative of the external function, it will work properly with  
> Bonmin as the solver in ampl.
>
> Thanks a lot for your time.
>
> Roy
>
>
> On 3/2/07, Roy Chen <ytboy0208 at gmail.com> wrote:
> Hello, Andreas
> I appreciate your timely and helpful response very much.
> I am kind of confused with the situation now because of the  
> following reasons:
> The way I need to define one of those integral functions in my  
> problem is like this:
> Integral(var1,var2,var3), assume I need these three inputs to do  
> the calculation.
> In my ampl .mod file, I used this function like this:
> set k;
> sum{s in k} Integral(x[s],y[s],z[s]) + .....; # x[s] is decision  
> variable, y[s] and z[s] are parameters defined in .dat file.
> The coded Integral() function will give you the value give those  
> three inputs.
> My question is:
> What are first and second order direvatives for in AMPL modeling?
> What kind of format you code the first and second order direvatives  
> for Bonmin?
> How Bonmin can know and use the first and second orfer direvatives  
> coded in AMPL external function?
> Again, if there could be a small example to illustrate this idea,  
> that's will be better, I think.
> Very appreciate your time.
> Best
> Roy
>
>
>
> On 3/2/07, Andreas Waechter < andreasw at watson.ibm.com> wrote:
> Hi Roy,
>
> > Could you actually provide one example which uses automatic  
> differentiation
> > to give the first and second direvatives with ampl external  
> function?
>
> I'm not sure, part of the code was not written by me and I'm not  
> sure if I
> can give it away.
>
> > Another question coming into my mind is:
> > Does that mean I don't need code to calculate the value of the  
> integration
> > function, instead, I just need to code to calculate the first and  
> second
> > direvatives?
>
> No, we still also need the values.
>
> However, if your function (including the integral) is such that the  
> first
> derivative is some analytic function without the integral, you  
> could just
> code the first derivatives using that analytic expression, from  
> which you
> could also easily the second derivative (as the first derivative of  
> that
> other function).
>
> > Again, if you can give me the example you mentioned in your  
> email, I would
> > appreciate it very much.
>
> Given what I wrote in the last paragraph, let me know if you still  
> want me
> to look for an example with ADOL-C.
>
> Cheers,
>
> Andreas
>
>
> _______________________________________________
> Bonmin mailing list
> Bonmin at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/bonmin

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://list.coin-or.org/pipermail/bonmin/attachments/20070304/d835311e/attachment.html


More information about the Bonmin mailing list