Hi,
In response to your query about convergence of the new model-free
optimisation protocol (full_analysis.py), I have added a function to
that script which will run the three tests - identical chi2, identical
model-free models, and identical diffusion tensor and model-free
parameter values. If you use this modified script, it will tell you
exactly when convergence has occured.
The improved 'full_analysis.py' script is currently only located
within the Subversion repository (http://svn.gna.org/viewcvs/relax/).
There are two ways to get this script. The first is to install svn
and checkout the stable 1.2 line from the repository. The other
option is to follow the given link, into the 1.2 repository line, and
then download it directly.
Cheers,
Edward
On 11/23/06, Sebastien Morin <[EMAIL PROTECTED]> wrote:
Hi !
For my model-free analysis, it finally converged.
However, when I first tried the final optimization, it didn't work
because of bug 7616 (I think)... I'm about to try that final
optimization again since I had to work on other things for the last few
weeks...
In summary, it took, for my 271 residue (220 residues with data at three
fields) :
1. local_tm : 15 hours on a Intel Pentium 4 CPU 3.20GHz (2MB L3cache)
2. sphere : 153h (~6 days) before convergence on a Intel Pentium 4 CPU
3.20GHz (2MB L3cache)
3. prolate : 244h (~10 days) before convergence on a Intel Xeon CPU 2.80GHz
4. oblate : 234h (~10 days) before convergence on a Intel Xeon CPU 2.80GHz
5. ellipsoid : 215h (~9 days) before convergence on a Intel Pentium 4
CPU 3.20GHz
Total (without the final optimization) with 4 CPU : ~10 days which is
better than I thought... (In fact, it converged before I noticed since I
wasn't looking at the results so much because I had several projects at
the same time).
Please see the file readme.txt to get detailed info about the CPUs and
also detailed times for the different optimization steps. As I told, it
converged before I stopped doing iterations and I made a lot of runs for
nothing especially for the sphere and ellipsoid diffusion tensors...
So, a question here, is looking at X2 convergence enough to say an
optimization converged or must I look at detailed individual parameters ?
Thanks !
Séb :)
Voici les fichiers input :
==========================
500_NOE.in.min
500_R1.in.min
500_R2.in.min
600_NOE.in.min
600_R1.in.min
600_R2.in.min
800_NOE.in.min
800_R1.in.min
800_R2.in.min
PSE-4_start22_rotate-pdbinertia_axial-600.pdb
Computers
===================
Pelican : 1 CPU -> Intel Pentium 4 CPU 3.20GHz
L1 data cache : 8KB
L2 unified cache : 512KB
L3 unified cache : 2MB
Pigeon : 1 CPU -> Intel Pentium 4 CPU 3.20GHz
L1 data cache : 8KB
L2 unified cache : 512KB
Hibou : 2 CPU -> Intel Xeon CPU 2.80GHz
L1 data cache : 8KB
L2 unified cache : 512KB
Modèle Round Machine Durée X2 Convergence ?
============ ===== ============= ======= ============= =============
local_tm --- Pelican 15h00
-----------------------------------------------
sphere MII init Pelican 0h10
1 Pelican 42h00 7718.74681973
2 Pelican 23h00 7533.93421105
3 Pelican 18h00 7512.54761496
4 Pelican 31h00 7514.7552298
5 Pelican 19h00 7515.04861607
6 Pelican 7h00 7525.64668195
7 Pelican 13h00 7516.27057309
8 Pelican 3h00 7516.27057309 Convergence
9 Pelican 3h00 7516.27057309 Convergence-2
10 Pelican 3h00 7516.27057309 Convergence-3
11 Pelican 3h00 7516.27057309 Convergence-4
12 Pelican 5h00 7516.27057309 Convergence-5
13 Pelican 3h00 7516.27057309 Convergence-6
14 Pelican 3h00 7516.27057309 Convergence-7
15 Pelican 3h00 7516.27057309 Convergence-8
16 Pelican 3h00 7516.27057309 Convergence-9
17 Pelican 5h00 7516.27057309 Convergence-10
18 Pelican 3h00 7516.27057309 Convergence-11
19 Pelican 3h00 7516.27057309 Convergence-12
20 Pelican 4h00 7516.27057309 Convergence-13
21 Pelican 4h00 7516.27057309 Convergence-14
prolate MIII init Hibou 0h20
1 Hibou 75h00 7337.50825218
2 Hibou 44h00 7116.88659287
3 Hibou 32h00 7102.52111138
4 Hibou 23h00 7081.39916081
5 Hibou 17h00 7084.09742106
6 Hibou 16h00 7089.93520425
7 Hibou 22h00 7088.84030659
8 Hibou 15h00 7086.72261677
9 Hibou 5h00 7086.72261677 Convergence
10 Hibou 3h00 7086.72261677 Convergence-2
11 Hibou 4h00 7086.72261677 Convergence-3
oblate MIV init Hibou 0h20
1 Hibou 75h00 7612.76117331
2 Hibou 44h00 7374.16115094
3 Hibou 23h00 7351.41096494
4 Hibou 40h00 7313.90788663
5 Hibou 19h00 7311.26394344
6 Hibou 12h00 7312.89980682
7 Hibou 21h00 7310.50446393
8 Hibou 6h00 7310.50446393 Convergence
9 Hibou 5h00 7310.50446393 Convergence-2
10 Hibou 5h00 7310.50446393 Convergence-3
ellipsoid MV init Pigeon 1h00
1 Pigeon 75h00 7318.54929865
2 Pigeon 31h00 7062.93677664
3 Pigeon 20h00 7044.01927992
4 Pigeon 21h00 7030.85159805
5 Pigeon 9h00 7027.99439236
6 Pigeon 18h00 7029.89513108
7 Pigeon 14h00 7027.60385085
8 Pigeon 14h00 7026.60491212
9 Pigeon 12h00 7022.72611396
10 Pigeon 4h00 7022.72611396 Convergence
11 Pigeon 4h00 7022.72611396 Convergence-2
12 Pigeon 4h00 7022.72611396 Convergence-3
13 Pigeon 4h00 7022.72611396 Convergence-4
14 Pigeon 4h00 7022.72611396 Convergence-5
15 Pigeon 5h00 7022.72611396 Convergence-6
16 Pigeon 4h00 7022.72611396 Convergence-7
17 Pigeon 4h00 7022.72611396 Convergence-8
-----------------------------------------------
final init Pelican short Failed because of bug 7616...
_______________________________________________
relax (http://nmr-relax.com)
This is the relax-users mailing list
[email protected]
To unsubscribe from this list, get a password
reminder, or change your subscription options,
visit the list information page at
https://mail.gna.org/listinfo/relax-users
