<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=us-ascii">
<meta name="Generator" content="Microsoft Exchange Server">
<!-- converted from text --><style><!-- .EmailQuote { margin-left: 1pt; padding-left: 4pt; border-left: #800000 2px solid; } --></style>
</head>
<body>
<meta content="text/html; charset=UTF-8">
<style type="text/css" style="">
<!--
p
{margin-top:0;
margin-bottom:0}
-->
</style>
<div dir="ltr">
<div id="x_divtagdefaultwrapper" dir="ltr" style="font-size:12pt; color:#000000; font-family:Calibri,Arial,Helvetica,sans-serif">
<p>From my personal experience, Minotaur's interface to both FilterSQP and IPOPT is very reliable, as are the local solvers themselves. I have cross-referenced the results for about 300 fairly large problems (up to 300 variables) and found no numerically significant
difference between the two. <br>
</p>
<p><br>
</p>
<p>One likely culprit for unexpected behavior in convergence but not the result, is that AMPL may have reformulated your problem. Try re-running AMPL with:</p>
<p><br>
</p>
<p>option presolve 0</p>
<p><br>
</p>
<p>and see if it makes a difference.<br>
</p>
<p><br>
</p>
<p>Always happy to fix bugs if you find one though!</p>
<p><br>
</p>
<p>--<br>
</p>
<p>Kind regards,</p>
<p>Dr Nikos Kazazakis, </p>
<p>Dept. of Chemical Engineering,<br>
</p>
<p>Imperial College London,</p>
<p>Prince Consort Road,</p>
<p>SW7 2AZ<br>
</p>
</div>
<hr tabindex="-1" style="display:inline-block; width:98%">
<div id="x_divRplyFwdMsg" dir="ltr"><font face="Calibri, sans-serif" color="#000000" style="font-size:11pt"><b>From:</b> minotaur-bounces@lists.mcs.anl.gov <minotaur-bounces@lists.mcs.anl.gov> on behalf of Roberto Chao <chaoroberto@gmail.com><br>
<b>Sent:</b> 06 December 2016 11:26:15<br>
<b>To:</b> Sven Leyffer<br>
<b>Cc:</b> minotaur@lists.mcs.anl.gov; Sven Leyffer<br>
<b>Subject:</b> Re: [Minotaur] FilterSQP engine strange behaviour</font>
<div> </div>
</div>
</div>
<font size="2"><span style="font-size:10pt;">
<div class="PlainText">Thankyou for your interest. We'll test Minotaur an filterSQP with more<br>
complicated optimization problems with known solution and I'll report<br>
you the results. We'll have to make some minor changes to Minotaur to<br>
avoid defining one function for evaluating every constraint because we<br>
have tens of nonlinear constraints.<br>
<br>
<br>
Thankyou<br>
Roberto Chao<br>
<br>
<br>
2016-12-05 16:15 GMT+01:00 Sven Leyffer <leyffer@anl.gov>:<br>
> Dear Roberto,<br>
><br>
> As I said earlier, I see nothing strange in the behavior. Different<br>
> algorithms will converge differently. The main point is that both converge<br>
> to the same limit point as predicted by their theory.<br>
><br>
> Kind regards,<br>
><br>
> Sven<br>
><br>
><br>
><br>
> On 12/04/2016 08:45 PM, Ashutosh Mahajan wrote:<br>
>><br>
>> Dear Roberto Chao<br>
>><br>
>> Sven Leyffer who developed the filter solver is the best person to answer<br>
>> your<br>
>> queries. Minotaur merely uses it as an external library.<br>
>><br>
>> --<br>
>> Regards<br>
>> Ashutosh Mahajan<br>
>> <a href="http://www.ieor.iitb.ac.in/amahajan">http://www.ieor.iitb.ac.in/amahajan</a><br>
>><br>
>> On Mon, Nov 28, 2016 at 04:00:58PM +0100, Roberto Chao wrote:<br>
>>><br>
>>> In this problem the SNQP method converges monotonically to the<br>
>>> solution more or less like ipopt engine does. The same cannot be said<br>
>>> of the minotaur SNQP implementation (filterSQP engine). It's true that<br>
>>> in the end it converges to the solution but this is a very easy<br>
>>> problem.<br>
>>><br>
>>> Regards<br>
>>> Roberto Chao<br>
>>><br>
>>> 2016-11-28 6:59 GMT+01:00 Ashutosh Mahajan <amahajan@iitb.ac.in>:<br>
>>>><br>
>>>> Dear Roberto<br>
>>>><br>
>>>> Both FilterSQP and Ipopt seem to be converging to the same point<br>
>>>> (x0, x1) = (0.707107, 0.707107) in roughly the same number of<br>
>>>> iterations. Can you be a bit more specific about what you feel is<br>
>>>> unexpected?<br>
>>>> Cheers.<br>
>>>><br>
>>>> --<br>
>>>> Regards<br>
>>>> Ashutosh Mahajan<br>
>>>> <a href="http://www.ieor.iitb.ac.in/amahajan">http://www.ieor.iitb.ac.in/amahajan</a><br>
>>>><br>
>>>> On Sun, Nov 27, 2016 at 08:38:10PM +0100, Roberto Chao wrote:<br>
>>>>><br>
>>>>> Hello, my name is Roberto. I'm an experienced C/C++/Java developer and<br>
>>>>> I'm currently very excited with the SNQP method and Minotaur seems to<br>
>>>>> me the perfect interface with filtersqp solver.<br>
>>>>><br>
>>>>> I've downloaded an successfully compiled minotaur-0.2.0 on a<br>
>>>>> Linux-x86_64 machine. The executable has successfully passed all unit<br>
>>>>> tests. I'd like to use minotaur to solve small-medium scale (tens of<br>
>>>>> constraints) non-linear programming problems through the SNQP method.<br>
>>>>><br>
>>>>> The first basic problem I've tried to solve has been taken from the<br>
>>>>> book: Practical Methods of Optimization (2nd edition) (Pag 296):<br>
>>>>><br>
>>>>> max x1+x2<br>
>>>>> st. x1^2+x2^2<=1<br>
>>>>><br>
>>>>> the FilterSQPEngine exhibits a strange behaviour:<br>
>>>>><br>
>>>>> filterSQP: version 20010817<br>
>>>>> (x0, x1) = (2, 0)<br>
>>>>> (x0, x1) = (1.25, 10)<br>
>>>>> (x0, x1) = (1.25, 5)<br>
>>>>> (x0, x1) = (2.06029, 2.24118)<br>
>>>>> (x0, x1) = (1.23322, 1.157)<br>
>>>>> (x0, x1) = (0.786654, 0.829407)<br>
>>>>> (x0, x1) = (0.720027, 0.707684)<br>
>>>>> (x0, x1) = (0.706927, 0.707408)<br>
>>>>> (x0, x1) = (0.707107, 0.707107)<br>
>>>>> FilterSQPEngine: total calls = 1<br>
>>>>> FilterSQPEngine: strong branching calls = 0<br>
>>>>> FilterSQPEngine: total time in solving = 0.003292<br>
>>>>> FilterSQPEngine: time in str branching = 0<br>
>>>>> FilterSQPEngine: total iterations = 8<br>
>>>>> FilterSQPEngine: strong br iterations = 0<br>
>>>>> solution status code = 1<br>
>>>>> solution status = ProvenLocalOptimal<br>
>>>>><br>
>>>>> However using IPopt engine gives me this result:<br>
>>>>><br>
>>>>> (x0, x1) = (2, 0)<br>
>>>>> (x0, x1) = (1.24837, 1.7)<br>
>>>>> (x0, x1) = (1.06331, 1.06783)<br>
>>>>> (x0, x1) = (0.831419, 0.831195)<br>
>>>>> (x0, x1) = (0.731859, 0.731869)<br>
>>>>> (x0, x1) = (0.708612, 0.708612)<br>
>>>>> (x0, x1) = (0.707108, 0.707108)<br>
>>>>> (x0, x1) = (0.707107, 0.707107)<br>
>>>>> Ipopt: total calls = 1<br>
>>>>> Ipopt: strong branching calls = 0<br>
>>>>> Ipopt: total time in solving = 0.047418<br>
>>>>> Ipopt: total time in presolve = 2e-06<br>
>>>>> Ipopt: time in str branching = 0<br>
>>>>> Ipopt: total iterations = 7<br>
>>>>> Ipopt: strong br iterations = 0<br>
>>>>> solution status code = 1<br>
>>>>> solution status = ProvenLocalOptimal<br>
>>>>><br>
>>>>> so I can assume I've correctly defined jacobian, hessian of lagrange,<br>
>>>>> objective an constraint evaluation callback methods.<br>
>>>>><br>
>>>>> Can you help me please?<br>
>>>>> Thank you in advance.<br>
>>>>> Roberto Chao<br>
>>>>> _______________________________________________<br>
>>>>> Minotaur mailing list<br>
>>>>> Minotaur@lists.mcs.anl.gov<br>
>>>>> <a href="https://lists.mcs.anl.gov/mailman/listinfo/minotaur">https://lists.mcs.anl.gov/mailman/listinfo/minotaur</a><br>
>>><br>
>>> _______________________________________________<br>
>>> Minotaur mailing list<br>
>>> Minotaur@lists.mcs.anl.gov<br>
>>> <a href="https://lists.mcs.anl.gov/mailman/listinfo/minotaur">https://lists.mcs.anl.gov/mailman/listinfo/minotaur</a><br>
><br>
><br>
> --<br>
> ---<br>
> Sven Leyffer<br>
> Senior Computational Mathematician<br>
> Argonne National Laboratory<br>
><br>
_______________________________________________<br>
Minotaur mailing list<br>
Minotaur@lists.mcs.anl.gov<br>
<a href="https://lists.mcs.anl.gov/mailman/listinfo/minotaur">https://lists.mcs.anl.gov/mailman/listinfo/minotaur</a><br>
</div>
</span></font>
</body>
</html>