|
»
»
Does sqp example work ?
|
View:
New views
20 Messages
—
Rating Filter:
Alert me
|
| < Prev | 1 - 2 - 3 | Next > |
 |
Does sqp example work ?
by Dupuis
::
Rate this Message:
Hello,
I'm trying to use sqp, and the example given in the help does not work. Instead, I get [x, obj, info, iter, nf, lambda] = sqp (x0, @phi, @g, []) x = -1.80000 1.70000 1.90000 -0.80000 -0.80000 obj = 0.020930 info = 101 iter = 1 nf = 44 lambda = 2.1964e-314 5.3050e-313 3.3471e-265 I tried setting the path to the strict minimum, i.e. path(genpath("/usr/share/octave/3.0.0/m")) addpath('/usr/lib/octave/3.0.0/oct/i486-pc-linux-gnu') then clearing everything, same result. Platform: PC with AMD Ahtlon, Debian: octave3.0/sid uptodate 3.0.0-1 (i386) Could someone check if the example is correct ? TIA Pascal Dupuis |
|
|
Does sqp example work ?
by John W. Eaton
::
Rate this Message:
On 11-Jan-2008, Dupuis wrote:
| I'm trying to use sqp, and the example given in the help does not work. | Instead, I get | [x, obj, info, iter, nf, lambda] = sqp (x0, @phi, @g, []) | x = | -1.80000 | 1.70000 | 1.90000 | -0.80000 | -0.80000 | | obj = 0.020930 | info = 101 | iter = 1 | nf = 44 | lambda = | | 2.1964e-314 | 5.3050e-313 | 3.3471e-265 | | I tried setting the path to the strict minimum, i.e. | path(genpath("/usr/share/octave/3.0.0/m")) | addpath('/usr/lib/octave/3.0.0/oct/i486-pc-linux-gnu') | then clearing everything, same result. | Platform: PC with AMD Ahtlon, Debian: octave3.0/sid uptodate 3.0.0-1 (i386) | | Could someone check if the example is correct ? It worked when I wrote the documentation because I just cut and pasted the output from an Octave session. jwe _______________________________________________ Bug-octave mailing list Bug-octave@... https://www.cae.wisc.edu/mailman/listinfo/bug-octave |
|
|
Re: Does sqp example work ?
by Søren Hauberg
::
Rate this Message:
fre, 11 01 2008 kl. 12:37 -0500, skrev John W. Eaton: > It worked when I wrote the documentation because I just cut and pasted > the output from an Octave session. I seem to remember writing this example, so I _think_ you can just blame me :-( Søren _______________________________________________ Bug-octave mailing list Bug-octave@... https://www.cae.wisc.edu/mailman/listinfo/bug-octave |
|
|
Does sqp example work ?
by John W. Eaton
::
Rate this Message:
On 11-Jan-2008, Dupuis wrote:
| Hello, | I'm trying to use sqp, and the example given in the help does not work. | Instead, I get | [x, obj, info, iter, nf, lambda] = sqp (x0, @phi, @g, []) | x = | -1.80000 | 1.70000 | 1.90000 | -0.80000 | -0.80000 | | obj = 0.020930 | info = 101 | iter = 1 | nf = 44 | lambda = | | 2.1964e-314 | 5.3050e-313 | 3.3471e-265 | | I tried setting the path to the strict minimum, i.e. | path(genpath("/usr/share/octave/3.0.0/m")) | addpath('/usr/lib/octave/3.0.0/oct/i486-pc-linux-gnu') | then clearing everything, same result. | Platform: PC with AMD Ahtlon, Debian: octave3.0/sid uptodate 3.0.0-1 (i386) | | Could someone check if the example is correct ? Please try the following patch. I think it would still be possible to avoid calculating Y only when it is really needed, but the reverse of the attached patch is not quite the way to do it. I also added a test to the sqp function based on the example in the doc string. Thanks, jwe src/ChangeLog: 2008-01-11 John W. Eaton <jwe@...> * DLD-FUNCTIONS/__qp__.cc (qp): Undo part of change from 2007-09-04 (compute Y at each iteration). scripts/ChangeLog: 2008-01-11 John W. Eaton <jwe@...> * optimization/sqp.m: New test from example in doc string. Index: src/DLD-FUNCTIONS/__qp__.cc =================================================================== RCS file: /cvs/octave/src/DLD-FUNCTIONS/__qp__.cc,v retrieving revision 1.14 diff -u -u -r1.14 __qp__.cc --- src/DLD-FUNCTIONS/__qp__.cc 17 Oct 2007 15:38:54 -0000 1.14 +++ src/DLD-FUNCTIONS/__qp__.cc 11 Jan 2008 18:13:55 -0000 @@ -226,6 +226,9 @@ // FIXME -- still remain to handle the case of // non-full rank active set matrix. + // Computing the Y matrix (orthogonal to Z) + Y = Aact.pseudo_inverse (); + // Reduced Hessian Matrix Zt = Z.transpose (); Matrix rH = Zt * H * Z; @@ -313,7 +316,6 @@ // Computing the multipliers only for the inequality // constraints that are active. We do NOT compute // multipliers for the equality constraints. - Y = Aact.pseudo_inverse (); Matrix Yt = Y.transpose (); Yt = Yt.extract_n (n_eq, 0, n_act-n_eq, n); lambda_tmp = Yt * (g + H * p); Index: scripts/optimization/sqp.m =================================================================== RCS file: /cvs/octave/scripts/optimization/sqp.m,v retrieving revision 1.13 diff -u -u -r1.13 sqp.m --- scripts/optimization/sqp.m 15 Oct 2007 15:30:04 -0000 1.13 +++ scripts/optimization/sqp.m 11 Jan 2008 18:13:52 -0000 @@ -712,3 +712,29 @@ endif ### endfunction + +%! function r = g (x) +%! +%! r = [ sumsq(x)-10; +%! x(2)*x(3)-5*x(4)*x(5); +%! x(1)^3+x(2)^3+1 ]; +%! endfunction +%! +%! function obj = phi (x) +%! obj = exp(prod(x)) - 0.5*(x(1)^3+x(2)^3+1)^2; +%! endfunction +%! +%!test +%! x0 = [-1.8; 1.7; 1.9; -0.8; -0.8]; +%! +%! [x, obj, info, iter, nf, lambda] = sqp (x0, @phi, @g, []) +%! +%! x_opt = [-1.717143501952599; +%! 1.595709610928535; +%! 1.827245880097156; +%! -0.763643103133572; +%! -0.763643068453300]; +%! +%! obj_opt = 0.0539498477702739 +%! +%! assert (all (abs (x-x_opt) < sqrt (eps)) && abs (obj-obj_opt) < sqrt (eps)); _______________________________________________ Bug-octave mailing list Bug-octave@... https://www.cae.wisc.edu/mailman/listinfo/bug-octave |
|
|
Re: Does sqp example work ?
by John W. Eaton
::
Rate this Message:
On 11-Jan-2008, Søren Hauberg wrote:
| I seem to remember writing this example, so I _think_ you can just blame | me :-( There's no need to assign blame here. I think I've fixed the bug. jwe _______________________________________________ Bug-octave mailing list Bug-octave@... https://www.cae.wisc.edu/mailman/listinfo/bug-octave |
|
|
Re: Does sqp example work ?
by Sergei Steshenko-2
::
Rate this Message:
--- "John W. Eaton" <jwe@...> wrote: [snip] > > I also added a test to the sqp function based on the example in the > doc string. > > Thanks, > > jwe Surprisingly, I was taught to add a test per bug found and fixed while I was young and green (opposed to ripe/seasoned) VLSI design and then verification idea. It was in pre-ATPG era. So, I fully support the idea to add a test per bug found and fixed - and I have already noticed many (fixed bugs, not added tests). Thanks, Sergei. Applications From Scratch: http://appsfromscratch.berlios.de/ ____________________________________________________________________________________ Be a better friend, newshound, and know-it-all with Yahoo! Mobile. Try it now. http://mobile.yahoo.com/;_ylt=Ahu06i62sR8HDtDypao8Wcj9tAcJ _______________________________________________ Bug-octave mailing list Bug-octave@... https://www.cae.wisc.edu/mailman/listinfo/bug-octave |
|
|
Re: Does sqp example work ?
by John W. Eaton
::
Rate this Message:
On 11-Jan-2008, Sergei Steshenko wrote:
| So, I fully support the idea to add a test per bug found and fixed - and I have | already noticed many (fixed bugs, not added tests). You are certainly welcome to submit patches that add tests. jwe _______________________________________________ Bug-octave mailing list Bug-octave@... https://www.cae.wisc.edu/mailman/listinfo/bug-octave |
|
|
Re: Does sqp example work ?
by Sergei Steshenko-2
::
Rate this Message:
--- "John W. Eaton" <jwe@...> wrote: > On 11-Jan-2008, Sergei Steshenko wrote: > > | So, I fully support the idea to add a test per bug found and fixed - and I have > | already noticed many (fixed bugs, not added tests). > > You are certainly welcome to submit patches that add tests. > > jwe > When you verify a bug fix, you enter some 'octave' code, don't you ? So, is it going to be this way: 1) you fix a bug; 2) you enter 'octave' code; 3) you copy-paste the into an Email message and send it to me; 4) I create an 'octave' file and send it back to you, the newly added file being the test; ? --Sergei. Applications From Scratch: http://appsfromscratch.berlios.de/ ____________________________________________________________________________________ Looking for last minute shopping deals? Find them fast with Yahoo! Search. http://tools.search.yahoo.com/newsearch/category.php?category=shopping _______________________________________________ Bug-octave mailing list Bug-octave@... https://www.cae.wisc.edu/mailman/listinfo/bug-octave |
|
|
Re: Does sqp example work ?
by dbateman3
::
Rate this Message:
Sergei Steshenko wrote:
> --- "John W. Eaton" <jwe@...> wrote: > >> On 11-Jan-2008, Sergei Steshenko wrote: >> >> | So, I fully support the idea to add a test per bug found and fixed - and I have >> | already noticed many (fixed bugs, not added tests). >> >> You are certainly welcome to submit patches that add tests. >> >> jwe >> > > When you verify a bug fix, you enter some 'octave' code, don't you ? > > So, is it going to be this way: > > 1) you fix a bug; > 2) you enter 'octave' code; > 3) you copy-paste the into an Email message and send it to me; > 4) I create an 'octave' file and send it back to you, the newly added file > being the test; > > ? Good god Sergei how old are you.. Perhaps you should learn to treat people with a bit of respect or if you can't just shut up.. D. _______________________________________________ Bug-octave mailing list Bug-octave@... https://www.cae.wisc.edu/mailman/listinfo/bug-octave |
|
|
Re: Does sqp example work ?
by Sergei Steshenko-2
::
Rate this Message:
--- David Bateman <adb014@...> wrote: > Sergei Steshenko wrote: > > --- "John W. Eaton" <jwe@...> wrote: > > > >> On 11-Jan-2008, Sergei Steshenko wrote: > >> > >> | So, I fully support the idea to add a test per bug found and fixed - and I have > >> | already noticed many (fixed bugs, not added tests). > >> > >> You are certainly welcome to submit patches that add tests. > >> > >> jwe > >> > > > > When you verify a bug fix, you enter some 'octave' code, don't you ? > > > > So, is it going to be this way: > > > > 1) you fix a bug; > > 2) you enter 'octave' code; > > 3) you copy-paste the into an Email message and send it to me; > > 4) I create an 'octave' file and send it back to you, the newly added file > > being the test; > > > > ? > > Good god Sergei how old are you.. Perhaps you should learn to treat > people with a bit of respect or if you can't just shut up.. > > D. > What is wrong with the procedure I described ? --Sergei. Applications From Scratch: http://appsfromscratch.berlios.de/ ____________________________________________________________________________________ Looking for last minute shopping deals? Find them fast with Yahoo! Search. http://tools.search.yahoo.com/newsearch/category.php?category=shopping _______________________________________________ Bug-octave mailing list Bug-octave@... https://www.cae.wisc.edu/mailman/listinfo/bug-octave |
|
|
Re: Does sqp example work ?
by dbateman3
::
Rate this Message:
Sergei Steshenko wrote:
> --- David Bateman <adb014@...> wrote: > >> Sergei Steshenko wrote: >>> --- "John W. Eaton" <jwe@...> wrote: >>> >>>> On 11-Jan-2008, Sergei Steshenko wrote: >>>> >>>> | So, I fully support the idea to add a test per bug found and fixed - and I have >>>> | already noticed many (fixed bugs, not added tests). >>>> >>>> You are certainly welcome to submit patches that add tests. >>>> >>>> jwe >>>> >>> When you verify a bug fix, you enter some 'octave' code, don't you ? >>> >>> So, is it going to be this way: >>> >>> 1) you fix a bug; >>> 2) you enter 'octave' code; >>> 3) you copy-paste the into an Email message and send it to me; >>> 4) I create an 'octave' file and send it back to you, the newly added file >>> being the test; >>> >>> ? >> Good god Sergei how old are you.. Perhaps you should learn to treat >> people with a bit of respect or if you can't just shut up.. >> >> D. >> > > What is wrong with the procedure I described ? > If you can't see the lack of respect that you have shown in the majority of e-mails that you have sent to this list (basically you treat us all as incompetents) then I don't think I can help. Please read http://math-atlas.sourceforge.net/faq.html#utone which applies equally to the Octave code, and then modify your future e-mails to the list accordingly and see how much more friendly people can be. D. _______________________________________________ Bug-octave mailing list Bug-octave@... https://www.cae.wisc.edu/mailman/listinfo/bug-octave |
|
|
Re: Does sqp example work ?
by Sergei Steshenko-2
::
Rate this Message:
--- David Bateman <adb014@...> wrote: > Sergei Steshenko wrote: > > --- David Bateman <adb014@...> wrote: > > > >> Sergei Steshenko wrote: > >>> --- "John W. Eaton" <jwe@...> wrote: > >>> > >>>> On 11-Jan-2008, Sergei Steshenko wrote: > >>>> > >>>> | So, I fully support the idea to add a test per bug found and fixed - and I have > >>>> | already noticed many (fixed bugs, not added tests). > >>>> > >>>> You are certainly welcome to submit patches that add tests. > >>>> > >>>> jwe > >>>> > >>> When you verify a bug fix, you enter some 'octave' code, don't you ? > >>> > >>> So, is it going to be this way: > >>> > >>> 1) you fix a bug; > >>> 2) you enter 'octave' code; > >>> 3) you copy-paste the into an Email message and send it to me; > >>> 4) I create an 'octave' file and send it back to you, the newly added file > >>> being the test; > >>> > >>> ? > >> Good god Sergei how old are you.. Perhaps you should learn to treat > >> people with a bit of respect or if you can't just shut up.. > >> > >> D. > >> > > > > What is wrong with the procedure I described ? > > > > If you can't see the lack of respect that you have shown in the majority > of e-mails that you have sent to this list (basically you treat us all > as incompetents) ??? When I suggested to accompany bug fixing with creating a (most likely short) testcase I expressed a methodology which I partly learned from others and partly developed myself. In the VLSI world it was customary for designers to develop basic tests for their own blocks; in fact both design code and the tests were delivered to the central tree during designers' release process. Since designer, basically, should not test his own code, rather, designer's claim that his/he code works should not be take too seriously, it was also customary for designers to write code for blocks of others, and there was also central verification team which wrote tests for the whole chip. Sometimes bugs were discovered in the end of, say, a one hour test, and for bug fixing/debugging such a long test is very inconvenient. So, designers whose blocks had bugs tried first to create a short test showing the same bug. Then, after fixing the bug, the short test was added to the pool of regression tests. But first the designer had to show that the bug was fixed. I.e. the point of all this is that typically designer already has tests showing bugs before the bugs are fixed and showing that the code works OK after the bugs are fixed. That is, the designers have these tests naturally, because of the nature of their work and accepted methodology. It looks to me that with 'octave' the situation is similar - code developers fix bugs using (typically short) pieces of 'octave' code, and that's why it looks natural to me to convert the pieces of code into regression tests. Regards, Sergei. Applications From Scratch: http://appsfromscratch.berlios.de/ ____________________________________________________________________________________ Be a better friend, newshound, and know-it-all with Yahoo! Mobile. Try it now. http://mobile.yahoo.com/;_ylt=Ahu06i62sR8HDtDypao8Wcj9tAcJ _______________________________________________ Bug-octave mailing list Bug-octave@... https://www.cae.wisc.edu/mailman/listinfo/bug-octave |
|
|
Re: Does sqp example work ?
by Thomas Weber-8
::
Rate this Message:
Am Sonntag, den 13.01.2008, 14:34 -0800 schrieb Sergei Steshenko: > --- David Bateman <adb014@...> wrote: > > > Sergei Steshenko wrote: > > > --- David Bateman <adb014@...> wrote: > > > > > >> Sergei Steshenko wrote: > > >>> --- "John W. Eaton" <jwe@...> wrote: > > >>> > > >>>> On 11-Jan-2008, Sergei Steshenko wrote: > > >>>> > > >>>> | So, I fully support the idea to add a test per bug found and fixed - and I have > > >>>> | already noticed many (fixed bugs, not added tests). > > >>>> > > >>>> You are certainly welcome to submit patches that add tests. > > >>>> > > >>>> jwe > > >>>> > > >>> When you verify a bug fix, you enter some 'octave' code, don't you ? > > >>> > > >>> So, is it going to be this way: > > >>> > > >>> 1) you fix a bug; > > >>> 2) you enter 'octave' code; > > >>> 3) you copy-paste the into an Email message and send it to me; > > >>> 4) I create an 'octave' file and send it back to you, the newly added file > > >>> being the test; > > >>> > > >>> ? > > >> Good god Sergei how old are you.. Perhaps you should learn to treat > > >> people with a bit of respect or if you can't just shut up.. > > >> > > >> D. > > >> > > > > > > What is wrong with the procedure I described ? With the procedure: nothing. In fact, I agree with you, that submitted bugs should be added as tests. That said, your e-mails come across as rude. I can't pinpoint exactly why, which probably means they aren't meant to be. The usual suspects (language barrier, lack of time when composing the mail, ...) might as well be responsible for that. > It looks to me that with 'octave' the situation is similar - code > developers fix bugs using (typically short) pieces of 'octave' code, > and that's why it looks natural to me to convert the pieces of code > into regression tests. You are right, so start doing it. And I really mean it. John and the other core devs are overworked, whereas stripping a test case from a submitted bug report is rather easy (some bugs are obviously an exception). Thomas _______________________________________________ Bug-octave mailing list Bug-octave@... https://www.cae.wisc.edu/mailman/listinfo/bug-octave |
|
|
Re: Does sqp example work ?
by Dupuis
::
Rate this Message:
I performed a first compilation, it failed. I hope I solved the problem, I'm retrying. Results tomorrow, if everything run well. Best regards Pascal |
|
|
Re: Does sqp example work ?
by Dupuis
::
Rate this Message:
Follow-up, and good news: with John's patch applied, sqp example works exactly as advertised. Thanks, John. Regards Pascal |
|
|
Re: Does sqp example work ?
by xmt
::
Rate this Message:
I've been using octave for a short while but I think it's a fantastic tool and I love it! I'm trying to use the SQP function for a marine engineering optimization problem, and I have the same problem: the SQP example doesn't seem to work. I'm using pre-built Octave 3.0 for windows (platform win. XP). How can I apply the patch that John has supplied for us? Thanks for your help! -xmt |
|
|
Re: Does sqp example work ?
by David Bateman
::
Rate this Message:
xmt wrote:
> Dupuis wrote: > >> Follow-up, and good news: with John's patch applied, sqp example works >> exactly as advertised. Thanks, John. >> >> Regards >> >> Pascal >> >> > > I've been using octave for a short while but I think it's a fantastic tool > and I love it! I'm trying to use the SQP function for a marine engineering > optimization problem, and I have the same problem: the SQP example doesn't > seem to work. I'm using pre-built Octave 3.0 for windows (platform win. XP). > How can I apply the patch that John has supplied for us? Thanks for your > help! > > -xmt > source file.. You'll need to wait for the first Octave 3.0.x bug fix release, unless you can rebuild it for yourself. D. -- David Bateman David.Bateman@... Motorola Labs - Paris +33 1 69 35 48 04 (Ph) Parc Les Algorithmes, Commune de St Aubin +33 6 72 01 06 33 (Mob) 91193 Gif-Sur-Yvette FRANCE +33 1 69 35 77 01 (Fax) The information contained in this communication has been classified as: [x] General Business Information [ ] Motorola Internal Use Only [ ] Motorola Confidential Proprietary _______________________________________________ Bug-octave mailing list Bug-octave@... https://www.cae.wisc.edu/mailman/listinfo/bug-octave |
|
|
Re: Does sqp example work ?
by Michael Goffioul-2
::
Rate this Message:
On Jan 28, 2008 10:23 AM, David Bateman <David.Bateman@...> wrote:
> You can't without a rebuild of the code as the patch modifies a C++ > source file.. You'll need to wait for the first Octave 3.0.x bug fix > release, unless you can rebuild it for yourself. As the patch involves an oct-file, it should be technically possible to compile only the __qp__.oct file using mkoctfile (provided that you have the MSVC 2005 compile environment installed on your computer). Michael. _______________________________________________ Bug-octave mailing list Bug-octave@... https://www.cae.wisc.edu/mailman/listinfo/bug-octave |
|
|
Re: Does sqp example work ?
by David Bateman
::
Rate this Message:
Michael Goffioul wrote:
> On Jan 28, 2008 10:23 AM, David Bateman <David.Bateman@...> wrote: > >> You can't without a rebuild of the code as the patch modifies a C++ >> source file.. You'll need to wait for the first Octave 3.0.x bug fix >> release, unless you can rebuild it for yourself. >> > > As the patch involves an oct-file, it should be technically possible > to compile only the __qp__.oct file using mkoctfile (provided that > you have the MSVC 2005 compile environment installed on your > computer). > > Michael. > > and so if he can build just __qp__.oct he'll he'll also be able to build Octave itself... Humm, seems that some time soon a 3.0.1 release will be needed.. D. -- David Bateman David.Bateman@... Motorola Labs - Paris +33 1 69 35 48 04 (Ph) Parc Les Algorithmes, Commune de St Aubin +33 6 72 01 06 33 (Mob) 91193 Gif-Sur-Yvette FRANCE +33 1 69 35 77 01 (Fax) The information contained in this communication has been classified as: [x] General Business Information [ ] Motorola Internal Use Only [ ] Motorola Confidential Proprietary _______________________________________________ Bug-octave mailing list Bug-octave@... https://www.cae.wisc.edu/mailman/listinfo/bug-octave |
|
|
Re: Does sqp example work ?
by John W. Eaton
::
Rate this Message:
On 28-Jan-2008, David Bateman wrote:
| Michael Goffioul wrote: | > On Jan 28, 2008 10:23 AM, David Bateman <David.Bateman@...> wrote: | > | >> You can't without a rebuild of the code as the patch modifies a C++ | >> source file.. You'll need to wait for the first Octave 3.0.x bug fix | >> release, unless you can rebuild it for yourself. | >> | > | > As the patch involves an oct-file, it should be technically possible | > to compile only the __qp__.oct file using mkoctfile (provided that | > you have the MSVC 2005 compile environment installed on your | > computer). | > | > Michael. | > | > | Yes, but the hurdle is the installation of MSVC 2005 build environment, | and so if he can build just __qp__.oct he'll he'll also be able to build | Octave itself... Humm, seems that some time soon a 3.0.1 release will be | needed.. I'm ready to do that if you think it is time. I've been applying most bug fixes to the 3.0 branch. If there are any fixes that have been applied only to cvs head, let me know and I'll try to apply them to the 3.0 branch as well. Are there any other pending problems that have not yet been fixed that we should address on the 3.0 branch? jwe _______________________________________________ Bug-octave mailing list Bug-octave@... https://www.cae.wisc.edu/mailman/listinfo/bug-octave |
| < Prev | 1 - 2 - 3 | Next > |
| Free embeddable forum powered by Nabble | Forum Help |
