Hmm not sure. I assume it is unregularised CGLS? You could try FISTA with f set to LeastSquares and g to Zero Function. That should solve the same problem as CGLS.
________________________________
From: Mailing list for software developers of the CCP in Tomographic Imaging <[log in to unmask]> on behalf of Gemma Fardell - UKRI STFC <[log in to unmask]>
Sent: 08 June 2020 16:12
To: [log in to unmask] <[log in to unmask]>
Subject: CGLS problem
Hi all,
I’m having problems with CGLS.
This is specifically the lego, Laminography dataset with the vector Astra projectors. But I’m not sure it’s a problem with the projectors (although I could easily be proved wrong!). The geometry itself is definitely ok as I get FDK to reconstruct, and I have the same problem with astra 1.9.9dev and 1.8.3.
CGLS seems to happily converge for 15 or so iterations (objective goes down, solution looks better), then it blows up.
You can see attached slice_c.tif (100 iterations, but it looked pretty consistent from 20 onwards) was the result I got last time I tried this (February maybe?) and the pdf is my current output after these 20 iterations now, after these 20 it morphs in to one featureless blur.
axpby was introduced since then, and I found and fixed a bug that was affecting it updating, but this happens if I don’t use that function too. Have there been any other changed what would affect this in the last few months? I can’t see anything specifically in CGLS.py, but maybe some deeper functionality?
Has anyone else seen a change with CGLS? I was having trouble with the Sophia beads dataset, so this could be the problem there too, but I haven’t investigated yet.
Iter Max Iter Time/Iter Objective
[s]
0 1000 0.000 3.42470e+07
1 1000 0.000 8.91496e+06
2 1000 114.129 4.09773e+06
3 1000 111.710 2.56780e+06
4 1000 121.898 1.70113e+06
5 1000 111.406 1.12592e+06
6 1000 111.262 7.90129e+05
7 1000 110.724 6.26495e+05
8 1000 111.337 4.96415e+05
9 1000 110.315 4.35742e+05
10 1000 111.503 4.50084e+05
11 1000 111.601 5.64865e+05
12 1000 110.762 7.02973e+05
13 1000 111.855 1.00553e+06
14 1000 110.926 1.49662e+06
15 1000 111.644 2.14943e+06
16 1000 111.985 3.27921e+06
17 1000 112.439 5.69564e+06
18 1000 110.871 8.97440e+06
19 1000 112.075 1.22793e+07
20 1000 108.565 2.04652e+07
I’ll take any suggestions, I’m banging my head against a brick wall here.
Gemma
This email and any attachments are intended solely for the use of the named recipients. If you are not the intended recipient you must not use, disclose, copy or distribute this email or any of its attachments and should notify the sender immediately and delete this email from your system. UK Research and Innovation (UKRI) has taken every reasonable precaution to minimise risk of this email or any attachments containing viruses or malware but the recipient should carry out its own virus and malware checks before opening the attachments. UKRI does not accept any liability for any losses or damages which the recipient may sustain due to presence of any viruses. Opinions, conclusions or other information in this message and attachments that are not related directly to UKRI business are solely those of the author and do not represent the views of UKRI.
________________________________
To unsubscribe from the CCPI-DEVEL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCPI-DEVEL&A=1
########################################################################
To unsubscribe from the CCPI-DEVEL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCPI-DEVEL&A=1
This message was issued to members of www.jiscmail.ac.uk/CCPI-DEVEL, a mailing list hosted by www.jiscmail.ac.uk, terms & conditions are available at https://www.jiscmail.ac.uk/policyandsecurity/
