[Insight-users] Problem with lbfgsb optimiser

[Tim Allman]

While doing some rigid registrations I found that occasionally the 
dreaded message "itk::ERROR: MeanSquaresImageToImageMetric(0x1caf8330): 
Too many samples map outside moving image buffer: 0 / 14986" would end 
it all. I could find nothing wrong with my code so I dived into the itk 
libraries with my debugger and discovered that code in lbfgsb.c failed 
occasionally and without apparent reason. I had thought that the 
registration had just lost its way but the trail of output parameters 
suggests not.

I don't have any more time to put into this because it could be in the 
logic of the Fortran code or the f2c translation (which seems to have 
been modified) and it would take too long to learn the logic of the 
code. I will just try another optimiser.

However, I have written a small program (attached) which demonstrates 
that problem. Perhaps it will be helpful to you.

Tim

-- 
Tim Allman, Ph.D.
35 Margaret Street,
Guelph Ontario N1E 5R6
Canada

-------------- next part --------------
A non-text attachment was scrubbed...
Name: lbfgsb-bug.tgz
Type: application/x-compressed-tar
Size: 4684 bytes
Desc: not available
URL: <http://www.itk.org/pipermail/insight-users/attachments/20130918/6da629a9/attachment.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: dr_tim_allman.vcf
Type: text/x-vcard
Size: 159 bytes
Desc: not available
URL: <http://www.itk.org/pipermail/insight-users/attachments/20130918/6da629a9/attachment.vcf>