Difference between revisions of "Data Challenges"

From CMB-S4 wiki
Jump to navigationJump to search
 
(62 intermediate revisions by 3 users not shown)
Line 1: Line 1:
=== Experiment Definitions ===
+
We separate "experiment definitions" (bands, resolutions, sensitivity level) from "sky models" (LCDM+dust+sync+).
 +
Each of these is designated by a number. The combination of a given experiment definition XX acting on a given sky model YY is designated XX.YY.
  
[Experiment Definitions]
+
Experiment models, sky models and the datasets obtained by applying for former to the latter are stored at NERSC at <br />
 +
'''/project/projectdirs/cmbs4/{expt_xx, sky_yy, data_xx.yy}'''
  
=== Input Sky Components ===
+
== Experiment Definitions ==
  
[Sky Components]
+
'''[[Experiment Definitions]]'''
  
=== Sky Models ===
+
== Input Sky Components ==
  
'''Sky model 00''' is simply LCDM + Gaussian dust + Gaussian Sync.
+
'''[[Sky Components]]'''
Since the dust and sync levels are set to the levels found in the BICEP/Keck field it wouldn't make any sense to use this model for a large sky fraction.
 
  
'''Sky Model 01''' is [https://arxiv.org/abs/1608.02841 PySM] run in a1d1f1s1 mode - i.e. with the default settings for AME, dust, free-free and synchrotron.
+
== Sky Models ==
  
'''Sky Model 02''' is [https://arxiv.org/abs/1608.02841 PySM] run in a2d4f1s3 mode - [https://github.com/bthorne93/PySM GitHub] for some description of what this means.
+
'''[[Sky Models]]'''
  
=== Sim Data Sets ===
 
  
'''01.00''' is the 01 experiment definition plus the 00 sky model.
+
== Sim Data Sets ==
  
=== NERSC Details ===
+
Many of the possible xx.yy combinations have been generated and are available on NERSC.
  
Shared space is available on [http://www.nersc.gov NERSC], where everyone is welcome to [http://crd.lbl.gov/cmb sign-up for an account] under the "Data Analysis for Post-Planck CMB Experiments" allocation (PI Borrill). The files are located in the shared CMB-S4 file space (/project/projectdirs/cmbs4). In order to manage this space, all files stored there should be owned by the cmbs4 project account but accessible to the cmbs4 group. To do this:
+
Most of 02.00 through 02.09 exist - these are the "Science Book config".
 +
 
 +
Most of 04.00 through 04.06 exist - these are the "CDT report config".
 +
 
 +
For experiment config 06, only foreground models 00, 07, and 09 were generated - these are the "DSR report config"
 +
 
 +
== NERSC Info ==
 +
 
 +
Shared space is available on [http://www.nersc.gov NERSC], where everyone is welcome to [http://crd.lbl.gov/cmb sign-up for an account] under the "Data Analysis for Post-Planck CMB Experiments" allocation (PI Borrill). The files are located in the shared CMB-S4 file space (<code>/global/cfs/cdirs/cmbs4</code>). In order to manage this space, all files stored there should be owned by the cmbs4 project account but accessible to the cmbs4 group. To do this:
 
* log in to the cmbs4 project account following the instructions [[http://www.nersc.gov/users/accounts/user-accounts/production-project-accounts here]]  
 
* log in to the cmbs4 project account following the instructions [[http://www.nersc.gov/users/accounts/user-accounts/production-project-accounts here]]  
 
* sync the files to an appropriate subdirectory in the project space
 
* sync the files to an appropriate subdirectory in the project space
Line 30: Line 38:
 
For any NERSC issues, including access to the filegroup and/or project account, please contact [mailto:jdborrill@lbl.gov Julian Borrill].
 
For any NERSC issues, including access to the filegroup and/or project account, please contact [mailto:jdborrill@lbl.gov Julian Borrill].
  
== Data Challenge 1 ==
+
=== HPSS Archive ===
Intended as a simple test to develop our practice, compare spectral- and map-based forecasting, contrast different map-based codes and methods, and coordinate r and non-r science forecasting.
 
 
 
'''[http://bicep.rc.fas.harvard.edu/cbischoff/20170210_data_challenge_1/index.html Maps pager for data challenge 1]''' (version 3, posted 2017-02-10)
 
 
 
* All maps are in Healpix format with nside = 512 and &#x2113;max = 1024.
 
* Maps have been filtered to remove all signal and noise below &#x2113; = 30.
 
 
 
=== Instrument specification ===
 
{| class="wikitable" style="text-align: right; margin-left: 51px;"
 
! Frequency (GHz) !!  30 !!  40 !!  85 !!  95 !! 145 !! 155 !! 220 !! 270
 
|-
 
| Bandwidth (GHz) || 9.0 || 12.0 || 20.4 || 22.8 || 31.9 || 34.1 || 48.4 || 59.4
 
|-
 
| Beam FWHM (arcmin) || 76.6 || 57.5 || 27.0 || 24.2 || 15.9 || 14.8 || 10.7 || 8.5
 
|-
 
|}
 
 
 
* Tophat bandpasses [[Tophat bands for Data Challenge | described in this posting]], added to above table --[[User:Cbischoff|Cbischoff]] ([[User talk:Cbischoff|talk]]) 15:51, 8 November 2016 (UTC)
 
* Beam widths copied from [http://users.physics.harvard.edu/~buza/20160531_fisher/ Victor's posting here] --[[User:Cbischoff|Cbischoff]] ([[User talk:Cbischoff|talk]]) 19:20, 14 December 2016 (UTC)
 
 
 
Beam window functions for each frequency can be found on NERSC in:
 
  /project/projectdirs/cmbs4/data_challenges/phase1/sky/wfunc/
 
In addition to beam smoothing at high ell, these window functions drop sharply to zero below ell=30 (maps for the data challenge contain no information below ell=30).
 
 
 
=== Masks: Clem ===
 
 
 
Noise inverse variance map is defined as a flat circular region with radius = 12&deg;, surrounded by a tapered region that falls to zero with cos profile over an additional 15 degrees. This map is defined to range from 0 (in unobserved pixels) to 1 (in the flat 12&deg; circular region); noise simulations are generated over the full sky from a C_l spectrum, then multiplied by 1 / sqrt(inverse variance) so that the noise amplitude blows up near the edge of the field.
 
 
 
* (Sum of inverse variance map) / (# of pixels in full sky) = 0.0293, so total noise power should be equivalent to fsky = 3%.
 
* Fraction of pixels that are observed is 5.4%.
 
 
 
File located on NERSC at:
 
  /project/projectdirs/cmbs4/data_challenges/phase1/sky/amask/map/fsky03_fall_n512.fits
 
 
 
=== Scalar, tensor & non-Gaussian CMB: Borrill ===
 
 
 
'''Lensed scalar CMB''', consistent with Planck 2015 and low &tau; parameters
 
* Lensed CMB maps are posted on NERSC in directory:
 
  /project/projectdirs/cmbs4/data_challenges/phase1/sky/llcdm/map/
 
 
 
'''Unlensed CMB'''
 
* Unlensed CMB maps are posted on NERSC in directory:
 
  /project/projectdirs/cmbs4/data_challenges/phase1/sky/ulcdm/map/
 
 
 
'''Corresponding Kappa Maps?'''
 
 
 
=== Extragalactic foregrounds: Alvarez / Battaglia / Bond / Stein===
 
 
 
Not included in Data Challenge 1
 
 
 
=== Galactic foregrounds ===
 
 
 
'''Gaussian dust'''
 
* Amplitude (in BB spectrum) is 4.25 &mu;K<sup>2</sup> at &nu; = 353 GHz, &#x2113; = 80. This is the value that was used for Science Book forecasts and is listed in [http://users.physics.harvard.edu/~buza/20160531_fisher/ Victor's 2016-05-31 posting] (Section 2). It also corresponds to the best fit dust amplitude from the [http://adsabs.harvard.edu/abs/2016PhRvL.116c1302B BK14 result].
 
* Dust amplitude in EE is 2x larger than BB. Dust amplitude in TT is 10x larger than EE (20x larger than BB). There is no TE, TB, or EB correlation.
 
* Dust scaling in frequency follows a greybody spectrum with &beta;<sub>dust</sub> = 1.6 and T<sub>dust</sub> = 19.6 K. This choice of parameters comes from [http://www.aanda.org/articles/aa/abs/2016/02/aa25034-14/aa25034-14.html PIPXXX].
 
* Dust D<sub>&#x2113;</sub> scaling follows a power law in ell with exponent &alpha;<sub>dust</sub> = -0.4. This parameter also comes from [http://www.aanda.org/articles/aa/abs/2016/02/aa25034-14/aa25034-14.html PIPXXX].
 
* Dust maps are posted on NERSC in directory:
 
  /project/projectdirs/cmbs4/data_challenges/phase1/sky/gdust/map/
 
 
 
'''Gaussian synchrotron'''
 
* Amplitude (in BB spectrum) is 3.8 &mu;K<sup>2</sup> at &nu; = 23 GHz, &#x2113; = 80. This is the value that was used for Science Book forecasts and is listed in [http://users.physics.harvard.edu/~buza/20160531_fisher/ Victor's 2016-05-31 posting] (Section 2). It also corresponds to the 95% upper limit from the [http://adsabs.harvard.edu/abs/2016PhRvL.116c1302B BK14 result].
 
* Sync amplitude in EE is 2x larger than BB. Sync amplitude in TT is 10x larger than EE (20x larger than BB). There is no TE, TB, or EB correlation.
 
* Sync scaling in frequency follows a power-law spectrum with &beta;<sub>sync</sub> = -3.1.
 
* Sync D<sub>&#x2113;</sub> scaling follows a power law in ell with exponent &alpha;<sub>sync</sub> = -0.6.
 
* Sync maps are posted on NERSC in directory:
 
  /project/projectdirs/cmbs4/data_challenges/phase1/sky/gsync/map/
 
 
 
=== Noise: homogeneous isotropic matched to N_ell: Victor / Clem / Colin / John K ===
 
 
 
'''White + 1/&#x2113; noise'''
 
* N<sub>&#x2113;</sub> functional form and parameters listed in [http://users.physics.harvard.edu/~buza/20161220_chkS4/ Victor's 2016-12-20 posting] (Table 2).
 
* Noise a<sub>&#x2113;m</sub> are calculated out to &#x2113; = 4096, but maps are rendered at nside = 512 with &#x2113;<sub>max</sub> = 1024. No beam smoothing or pixel window function is applied.
 
* Full sky noise maps are generated from N<sub>&#x2113;</sub>, then scaled by the square root of variance map to produce noise that is lowest in the center of the observed field and blows up towards the edge.
 
* Noise maps are posted on NERSC in director:
 
  /project/projectdirs/cmbs4/data_challenges/phase1/sky/noise/map/
 
* Maps with filename like <tt>r0000_f030_n512.fits</tt> are full sky noise maps. Filenames like <tt>r0000_f030_n512_b00p0_mfsky03.fits</tt> have been multiplied by the square root of variance map.
 
 
 
=== Combined noise + CMB + dust + sync maps ===
 
* Four versions of these maps, with different levels of residual lensing signal. Example file names are:
 
** A<sub>L</sub> = 1 : <tt>r0000_f030_AL1p0_n512_mfsky03.fits</tt>
 
** A<sub>L</sub> = 0.3 : <tt>r0000_f030_AL0p3_n512_mfsky03.fits</tt>
 
** A<sub>L</sub> = 0.1 : <tt>r0000_f030_AL0p1_n512_mfsky03.fits</tt>
 
** A<sub>L</sub> = 0.03 : <tt>r0000_f030_AL0p03_n512_mfsky03.fits</tt>
 
* Maps with partial lensing are constructed by the following combination:
 
  partially_lensed_map = sqrt(A_L) * lensed_map + (1 - sqrt(A_L)) * unlensed_map
 
* Maps are posted on NERSC in directory:
 
  /project/projectdirs/cmbs4/data_challenges/phase1/sky/comb/map/
 
 
 
=== Phase 1 Analysis ===
 
  
Please add your main results here and link to your more detailed posting in the usual logbook. In the logbook, please specify which maps you made use of in your analysis, i.e. whether you used the combined maps, CMB simulations, noise simulations, etc.
+
Older data challenge maps are moved to the [https://docs.nersc.gov/filesystems/archive/ High Performance Storage System]. These are stored in the <code>low_ell_BB</code> directory in the <code>cmbs4</code> user home directory of HPSS. For each Data Challenge, maps are broken up in several different archives to achieve archive file sizes of ~few 100 GB.
  
=== Schedule ===
+
List the data challenge sets that have been archived here:
* Dec 7: Beta version of Phase 1 maps
+
02.00,  02.01, 02.02, 02.03, 02.04, 02.05, 02.06, 02.09
* Dec 14: Telecon
+
02b.00, 02b.03
* Dec 17: Noise power (N_ell) prescriptions to Clem from Victor and Colin
+
02c.00, 02c.03
* Dec 21: A few realizations of Phase 1 maps delivered by Clem
+
03.00,  03.03
* Dec 21: Telecon (note phase shift)
+
03b.00, 03b.03
* Jan 4: Telecon
+
03c.00, 03c.03
* Jan 12: 100 realizations of Phase 1 maps posted to NERSC
 
* Jan 18: Telecon
 
* Feb 1: Telecon
 
* Feb 15: Telecon
 
* Feb 27, 28, Mar 1: S4 general meeting at SLAC
 
* Mar 2, 3: S4 CDT meeting at SLAC
 

Latest revision as of 11:39, 25 February 2020

We separate "experiment definitions" (bands, resolutions, sensitivity level) from "sky models" (LCDM+dust+sync+). Each of these is designated by a number. The combination of a given experiment definition XX acting on a given sky model YY is designated XX.YY.

Experiment models, sky models and the datasets obtained by applying for former to the latter are stored at NERSC at
/project/projectdirs/cmbs4/{expt_xx, sky_yy, data_xx.yy}

Experiment Definitions

Experiment Definitions

Input Sky Components

Sky Components

Sky Models

Sky Models


Sim Data Sets

Many of the possible xx.yy combinations have been generated and are available on NERSC.

Most of 02.00 through 02.09 exist - these are the "Science Book config".

Most of 04.00 through 04.06 exist - these are the "CDT report config".

For experiment config 06, only foreground models 00, 07, and 09 were generated - these are the "DSR report config"

NERSC Info

Shared space is available on NERSC, where everyone is welcome to sign-up for an account under the "Data Analysis for Post-Planck CMB Experiments" allocation (PI Borrill). The files are located in the shared CMB-S4 file space (/global/cfs/cdirs/cmbs4). In order to manage this space, all files stored there should be owned by the cmbs4 project account but accessible to the cmbs4 group. To do this:

  • log in to the cmbs4 project account following the instructions [here]
  • sync the files to an appropriate subdirectory in the project space
  • ensure that the permissions are set appropriately (g+rX,o-rwx)

Remember to include a README and to post the details on this wiki page.

For any NERSC issues, including access to the filegroup and/or project account, please contact Julian Borrill.

HPSS Archive

Older data challenge maps are moved to the High Performance Storage System. These are stored in the low_ell_BB directory in the cmbs4 user home directory of HPSS. For each Data Challenge, maps are broken up in several different archives to achieve archive file sizes of ~few 100 GB.

List the data challenge sets that have been archived here:

02.00,  02.01, 02.02, 02.03, 02.04, 02.05, 02.06, 02.09
02b.00, 02b.03
02c.00, 02c.03
03.00,  03.03
03b.00, 03b.03
03c.00, 03c.03