[CosmoMC] Error in Matrix_Diagonalize

Use of Cobaya. camb, CLASS, cosmomc, compilers, etc.
Post Reply
Su GuanJhen
Posts: 8
Joined: November 28 2015
Affiliation: National Taiwan University

[CosmoMC] Error in Matrix_Diagonalize

Post by Su GuanJhen » April 27 2016

I want to calculate BICEP2/Keck array likelihood by using CosmoMC. I set action=0 to do the Monte-Carlo. However, I get the error message like this:

Code: Select all

Number of MPI processes:           1
 file_root:BK14_baseline
 Random seeds:  3170, 11176 rand_inst:   1
 Doing non-linear Pk: F
 Doing CMB lensing: T
 Doing non-linear lensing: T
 TT lmax =  2500
 EE lmax =  2500
 ET lmax =  2500
 BB lmax =  2500
 PP lmax =  2500
 lmax_computed_cl  =  2500
 Computing tensors: T
 lmax_tensor    =   600
 max_eta_k         =    14000.0000    
 transfer kmax     =    5.00000000    
 adding parameters for: BK14_dust
 Fast divided into            1  blocks
  8 parameters ( 1 slow ( 1 semi-slow),  7 fast ( 0 semi-fast))
 starting Monte-Carlo
 Error in Matrix_Diagonalize
 MpiStop:            0

I have used Monte-Carlo method with Planck likelihood and BKPlanck likelihood before, and done it successfully. I have no idea why it doesn't work with BICEP2/Keck array likelihood.

Can anyone tell me what "error in matrix_diagonalize" refer to?
Thank you.

ps: The following is the code from BK14_01_baseline.ini, which is used to calculate the likelihood with BICEP2/Keck array data.

Code: Select all

# BK14 foreground and data selection settings for the fiducial analysis
DEFAULT(batch2/BK14only.ini)
# General settings
DEFAULT(batch2/common.ini)

# Root name for files produced
file_root = BK14_baseline

# Folder where files (chains, checkpoints, etc.) are stored
#root_dir = chains/BK14_01_baseline/
root_dir = chains/

# From baseline analysis run
propose_matrix = planck_covmats/BK14.covmat

# Convergence limits
MPI_Converge_Stop = 0.001
MPI_Limit_Converge_Err = 0.1

# Run chains.
action = 0

Antony Lewis
Posts: 1941
Joined: September 23 2004
Affiliation: University of Sussex
Contact:

Re: [CosmoMC] Error in Matrix_Diagonalize

Post by Antony Lewis » April 27 2016

May be a problem with your covmat planck_covmats/BK14.covmat ?

Su GuanJhen
Posts: 8
Joined: November 28 2015
Affiliation: National Taiwan University

[CosmoMC] Error in Matrix_Diagonalize

Post by Su GuanJhen » April 28 2016

It seems that the problem is not from BK14.covmat because I comment out this line

Code: Select all

propose_matrix = planck_covmats/BK14.covmat 
but the same error occurs again.

I think *.covmat file is used for speeding up the convergence of chains, so comment out this line is OK. Am I right?

ps: The following is BK14.covmat

Code: Select all

# r BBdust BBsync BBalphadust BBbetadust BBalphasync BBbetasync BBdustsynccorr
  7.5425802E-04  -1.8946221E-03  -1.2781294E-03  -3.7511887E-04   5.5909196E-04   7.1333991E-06  -4.9113465E-04  -3.3723225E-05
 -1.8946221E-03   1.2286468E+00   1.1709482E-02  -1.1354681E-01   4.9949561E-02  -1.4107626E-03  -2.5995176E-03  -2.4261615E-04
 -1.2781294E-03   1.1709482E-02   1.3764452E+00  -3.3488097E-03  -1.1247582E-03   1.1293779E-02  -3.2439708E-02  -4.9098566E-02
 -3.7511887E-04  -1.1354681E-01  -3.3488097E-03   6.7573749E-02  -1.0507401E-03   2.7746148E-04  -8.9363325E-04  -9.1573147E-04
  5.5909196E-04   4.9949561E-02  -1.1247582E-03  -1.0507401E-03   1.0177810E-02   8.5427733E-05  -4.6402462E-05   1.4047561E-04
  7.1333991E-06  -1.4107626E-03   1.1293779E-02   2.7746148E-04   8.5427733E-05   8.1394032E-02  -2.2922411E-04  -7.8329110E-04
 -4.9113465E-04  -2.5995176E-03  -3.2439708E-02  -8.9363325E-04  -4.6402462E-05  -2.2922411E-04   8.0185145E-02  -2.4264992E-03
 -3.3723225E-05  -2.4261615E-04  -4.9098566E-02  -9.1573147E-04   1.4047561E-04  -7.8329110E-04  -2.4264992E-03   2.2036496E-02

Antony Lewis
Posts: 1941
Joined: September 23 2004
Affiliation: University of Sussex
Contact:

Re: [CosmoMC] Error in Matrix_Diagonalize

Post by Antony Lewis » April 29 2016

In that case it's probably in the BKP likelihood. Check the supplied test.ini runs OK, if it does, then may be specific to BK14_baseline.

Post Reply