cosmoMC - Reach the Gelman-Rubin limit - where defined

Use of Cobaya. camb, CLASS, cosmomc, compilers, etc.
Post Reply
Dournac Fabien
Posts: 74
Joined: May 18 2019
Affiliation: IRAP
Contact:

cosmoMC - Reach the Gelman-Rubin limit - where defined

Post by Dournac Fabien » February 16 2022

Hello,

I make running the Planck case (test_planck.ini) for MCMC with cosmoMC. I have launched it with 4 MPI processes.

Here a part of output during the run :
Chain 4 MPI communicating
Chain 3 MPI communicating
Chain 2 MPI communicating
Chain 1 MPI communicating
Current convergence R-1 = 2.2743432E-02 chain steps = 19045
slow changes 17153 power changes 0
updating proposal density
Chain:2 drag accpt: 0.4110676 fast/slow 130.0940 slow: 17059
Chain:0 drag accpt: 0.4070620 fast/slow 129.9818 slow: 17190
Chain:3 drag accpt: 0.4085876 fast/slow 129.8564 slow: 17459
Chain:2 drag accpt: 0.4113056 fast/slow 130.0898 slow: 17117
Chain:1 drag accpt: 0.4193379 fast/slow 129.5778 slow: 17304
Chain:0 drag accpt: 0.4066948 fast/slow 129.9687 slow: 17270
Chain:3 drag accpt: 0.4084697 fast/slow 129.8438 slow: 17532
Chain:1 drag accpt: 0.4192181 fast/slow 129.5793 slow: 17368
Chain:2 drag accpt: 0.4107864 fast/slow 130.0641 slow: 17202
Chain:0 drag accpt: 0.4068378 fast/slow 129.9641 slow: 17335
Chain:3 drag accpt: 0.4082053 fast/slow 129.8123 slow: 17602
Chain:2 drag accpt: 0.4110449 fast/slow 130.0597 slow: 17260
Chain:1 drag accpt: 0.4191867 fast/slow 129.5757 slow: 17436
Chain:0 drag accpt: 0.4067691 fast/slow 129.9615 slow: 17406
Chain:3 drag accpt: 0.4081947 fast/slow 129.8027 slow: 17666
Chain:1 drag accpt: 0.4194605 fast/slow 129.5664 slow: 17491
Chain:2 drag accpt: 0.4108091 fast/slow 130.0551 slow: 17339
Chain:0 drag accpt: 0.4069947 fast/slow 129.9479 slow: 17463
Chain:3 drag accpt: 0.4081841 fast/slow 129.8006 slow: 17731
Chain:1 drag accpt: 0.4194066 fast/slow 129.5679 slow: 17556
Chain:2 drag accpt: 0.4103414 fast/slow 130.0246 slow: 17421
Chain:0 drag accpt: 0.4072398 fast/slow 129.9531 slow: 17522
Chain:1 drag accpt: 0.4197212 fast/slow 129.5574 slow: 17610
Chain:3 drag accpt: 0.4081321 fast/slow 129.7970 slow: 17800
Chain:2 drag accpt: 0.4103650 fast/slow 130.0158 slow: 17488
Chain:0 drag accpt: 0.4072120 fast/slow 129.9432 slow: 17591
Chain:1 drag accpt: 0.4197747 fast/slow 129.5678 slow: 17671
Chain:3 drag accpt: 0.4083291 fast/slow 129.7956 slow: 17856
Chain:0 drag accpt: 0.4073299 fast/slow 129.9486 slow: 17658
Chain:1 drag accpt: 0.4201508 fast/slow 129.5688 slow: 17720
Chain:3 drag accpt: 0.4088353 fast/slow 129.7996 slow: 17905
Chain:2 drag accpt: 0.4101562 fast/slow 130.0092 slow: 17566
Chain:0 drag accpt: 0.4073848 fast/slow 129.9396 slow: 17726
Chain:3 drag accpt: 0.4088019 fast/slow 129.7966 slow: 17974
Chain:1 drag accpt: 0.4200521 fast/slow 129.5692 slow: 17792
Chain:2 drag accpt: 0.4098445 fast/slow 130.0008 slow: 17646
Chain:3 drag accpt: 0.4089537 fast/slow 129.7883 slow: 18038
Chain:0 drag accpt: 0.4069032 fast/slow 129.9253 slow: 17812
Chain:1 drag accpt: 0.4199756 fast/slow 129.5783 slow: 17861
Chain:2 drag accpt: 0.4098277 fast/slow 129.9987 slow: 17713
Chain:1 drag accpt: 0.4200274 fast/slow 129.5741 slow: 17919
Chain:3 drag accpt: 0.4090226 fast/slow 129.7858 slow: 18105
Chain:0 drag accpt: 0.4066925 fast/slow 129.9309 slow: 17892
Chain:2 drag accpt: 0.4092292 fast/slow 129.9836 slow: 17801
Chain:1 drag accpt: 0.4202913 fast/slow 129.5749 slow: 17974
Chain:3 drag accpt: 0.4089888 fast/slow 129.7732 slow: 18172
Chain:0 drag accpt: 0.4068716 fast/slow 129.9373 slow: 17950
Chain:2 drag accpt: 0.4093597 fast/slow 129.9755 slow: 17863
Chain:1 drag accpt: 0.4203629 fast/slow 129.5689 slow: 18033
Chain:3 drag accpt: 0.4090570 fast/slow 129.7808 slow: 18236
Chain:0 drag accpt: 0.4069680 fast/slow 129.9294 slow: 18009
Chain:1 drag accpt: 0.4204129 fast/slow 129.5579 slow: 18099
Chain:2 drag accpt: 0.4089332 fast/slow 129.9864 slow: 17956
Chain:3 drag accpt: 0.4092465 fast/slow 129.7694 slow: 18293
Chain:0 drag accpt: 0.4068608 fast/slow 129.9281 slow: 18077
Chain:2 drag accpt: 0.4088996 fast/slow 129.9767 slow: 18019
Chain:1 drag accpt: 0.4201681 fast/slow 129.5512 slow: 18177
Chain:0 drag accpt: 0.4068352 fast/slow 129.9205 slow: 18127
Chain:3 drag accpt: 0.4089293 fast/slow 129.7742 slow: 18372
Chain:1 drag accpt: 0.4203860 fast/slow 129.5433 slow: 18231
Chain:2 drag accpt: 0.4089071 fast/slow 129.9722 slow: 18085
Chain:0 drag accpt: 0.4067696 fast/slow 129.9097 slow: 18196
Chain:3 drag accpt: 0.4093394 fast/slow 129.7684 slow: 18423
Chain:1 drag accpt: 0.4203936 fast/slow 129.5242 slow: 18295
Chain:2 drag accpt: 0.4087315 fast/slow 129.9514 slow: 18150
Chain:3 drag accpt: 0.4094859 fast/slow 129.7591 slow: 18482
Chain:0 drag accpt: 0.4065842 fast/slow 129.9055 slow: 18269
Chain:1 drag accpt: 0.4202970 fast/slow 129.5176 slow: 18366
Chain:3 drag accpt: 0.4097117 fast/slow 129.7604 slow: 18534
Chain:2 drag accpt: 0.4085170 fast/slow 129.9410 slow: 18214
Chain:0 drag accpt: 0.4066398 fast/slow 129.9004 slow: 18334
Chain 2 MPI communicating
Chain 4 MPI communicating
Chain 3 MPI communicating
Chain 1 MPI communicating
Current convergence R-1 = 2.6739364E-02 chain steps = 20406
slow changes 18372 power changes 0
updating proposal density
Chain:1 drag accpt: 0.4204293 fast/slow 129.5131 slow: 18429
i just want to know where is defined the R-1 quantity value : it seems to be the Gelman-Rubin criterion but I can't find it in input files, doesn't it ?

Here is the content of test_planck.ini :
#Check that the Planck likelihood files run OK, see readme_planck.html to set up
#Planck 2018, default just include native likelihoods (others require clik)
DEFAULT(batch3/plik_rd12_HM_v22_TTTEEE.ini)
DEFAULT(batch3/lowl.ini)

DEFAULT(batch3/simall_EE.ini)

#BAO compilation
DEFAULT(batch3/BAO.ini)

#Other settings same as test.ini
DEFAULT(test.ini)

#Uncomment if you want to include priors even when parameters fixed
#include_fixed_parameter_priors = T

#expected result for -(log like)
test_check_compare = 1820.775
By the way, what does "test_check_compare" mean ? Is it about [math] ?

Regards.

Antony Lewis
Posts: 1936
Joined: September 23 2004
Affiliation: University of Sussex
Contact:

Re: cosmoMC - Reach the Gelman-Rubin limit - where defined

Post by Antony Lewis » February 16 2022

See the

Code: Select all

MPI_Converge_Stop
etc. parameters in batch3/common.ini.

The compare number is for testing a single likelihood with action=4 (e.g. for travis unit tests).

Dournac Fabien
Posts: 74
Joined: May 18 2019
Affiliation: IRAP
Contact:

Re: cosmoMC - Reach the Gelman-Rubin limit - where defined

Post by Dournac Fabien » February 17 2022

Hello,

sorry but could you explain please a little bit better which corresponds to a "travis unit tests" ? Is it related to Gelman-Rubin criterion ?

Best regards

Antony Lewis
Posts: 1936
Joined: September 23 2004
Affiliation: University of Sussex
Contact:

Re: cosmoMC - Reach the Gelman-Rubin limit - where defined

Post by Antony Lewis » February 17 2022

Code: Select all

test_check_compare
is only used for action=4 for testing.

Post Reply