[ITK-users] Problem with ITK eigensystem

[Olivier Commowick]

Oh, and I remember only now : in fact it might just be that applying the patch I sent in the email solves this problem of infinite loop while still producing nan values

> On 2 Feb 2017, at 09:27, Olivier Commowick <Olivier.Commowick at inria.fr> wrote:
> 
> Hi,
> 
> Thanks for your test. There might be a bug in my test or the way you ran it since for me it doesn’t take a few minutes but rather less than a second. That being said since it produces nans it might be that on some OS it goes into an infinite loop...
> 
> Olivier
> 
>> On 2 Feb 2017, at 00:02, Dženan Zukić <dzenanz at gmail.com <mailto:dzenanz at gmail.com>> wrote:
>> 
>> Hi Olivier,
>> 
>> I tried running you example. One thing struck me right away: ITK version is much slower (after a few minutes running in release mod I killed it) than VNL version (about a second in debug mode) for your input.
>> 
>> I can't comment the patch content or the numeric stability. Perhaps Brad and Hans can pitch in?
>> 
>> Regards,
>> Dženan
>> 
>> On Tue, Jan 31, 2017 at 11:24 AM, Olivier Commowick <Olivier.Commowick at inria.fr <mailto:Olivier.Commowick at inria.fr>> wrote:
>> Hello,
>> 
>> I think I have found a bug in the behaviour of ITK’s symmetric eigensystem. In the attached link (http://olivier.commowick.org/tmp/Test.zip <http://olivier.commowick.org/tmp/Test.zip>) you will find an example code and matrix that fails with it and not with VNL’s symmetric eigensystem (run on macOS 10.12, Xcode 8.2.1).
>> 
>> It seems to me that there are subtle differences in between ITK and VNL versions, VNL being more robust (and the same as what Matlab gives, I didn’t test python). Here is what I have found so far :
>> 
>> - the hypothenuse function (hypot for ITK and pythag for VNL) do not get always the exact same result up to a very far away decimal
>> - some loop indices were wrong I think in the ITK eigensystem. Some do vary in the VNL version, not in ITK’s equivalent
>> 
>> In the patch file attached, I have made a set of corrections that make things in ITK eigensystem closer to VNL but still the ITK version fails. I have to say I am getting completely lost here so I am calling upon the community for help. If anyone knows this enough, could you try the test attached and see if you have any ideas of what is going on. That would help us a lot.
>> 
>> The test attached can be compiled with cmake and run with the text file included : ./test_eigensystem data.txt 
>> 
>> Thanks in advance
>> Olivier
>> 
>> 
>> 
>> ---
>> 
>> Olivier Commowick, Ph.D.
>> Research Scientist
>> INRIA Rennes - Bretagne Atlantique, VISAGES Team
>> Campus de Beaulieu
>> 35042 Rennes
>> FRANCE
>> 
>> Phone: +33 2 99 84 25 92 <tel:+33%202%2099%2084%2025%2092>
>> Email: Olivier.Commowick at inria.fr <mailto:Olivier.Commowick at inria.fr>
>> Web: http://olivier.commowick.org/ <http://olivier.commowick.org/>
>> 
>> 
>> 
>> 
>> 
>> 
>> _____________________________________
>> Powered by www.kitware.com <http://www.kitware.com/>
>> 
>> Visit other Kitware open-source projects at
>> http://www.kitware.com/opensource/opensource.html <http://www.kitware.com/opensource/opensource.html>
>> 
>> Kitware offers ITK Training Courses, for more information visit:
>> http://www.kitware.com/products/protraining.php <http://www.kitware.com/products/protraining.php>
>> 
>> Please keep messages on-topic and check the ITK FAQ at:
>> http://www.itk.org/Wiki/ITK_FAQ <http://www.itk.org/Wiki/ITK_FAQ>
>> 
>> Follow this link to subscribe/unsubscribe:
>> http://public.kitware.com/mailman/listinfo/insight-users <http://public.kitware.com/mailman/listinfo/insight-users>
>> 
>> 
> 
> 
> _____________________________________
> Powered by www.kitware.com
> 
> Visit other Kitware open-source projects at
> http://www.kitware.com/opensource/opensource.html
> 
> Kitware offers ITK Training Courses, for more information visit:
> http://www.kitware.com/products/protraining.php
> 
> Please keep messages on-topic and check the ITK FAQ at:
> http://www.itk.org/Wiki/ITK_FAQ
> 
> Follow this link to subscribe/unsubscribe:
> http://public.kitware.com/mailman/listinfo/insight-users


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/insight-users/attachments/20170202/1c768715/attachment.html>