E-EMC2
Principled Poisson Image "Deconvolution"/"Reconstruction" with Error Bars + Significance Tests
NEW!! E-EMC2 Dec 2008 Simpler test runs, associated small python utilities. A. Connors
These are test files for the principled image deconvolution, fitting,
and feature-detection/extraction method based on Esch et. al.
The code can be obtained from
David van Dyk at University of California,
Irvine; or
Nathan Stein
at the Harvard Statistics Department.
Thanks to Nathan Stein for sugesting this simpler test format.
The test image is 8x8, with a 5x3 "E" within it, on a constant background.
If your set-up is working, you should see the "E" clearly in the mean
and less clearly in the individual MCMC samples , when using a flat
or no background. Nothing significant should be visible
when using the 'true' null model as background.
The gzipped tar file is
HERE .
"gunzip" it, extract the files, and read the AAA_EEMC2_READ.ME file.
Please let us know (aconnors AT eurekabayes.com) if you have
questions or problems.
EMC2 Dec 2006 Version test runs, associated small python utilities. A. Connors
- Example files used -- models and "datons" (i.e. simulated Poisson
counts; term due to J. Lochner) -- in Connors and van Dyk's SCMA4
paper. They are in "R"-readable text format, padded to be square pwers
of two (128x128). They are based on EGRET >1GeV all-sky daya,all
viewing periods. Appreciation to HEASARC, the Compton GRO and EGRET
teams. The diffuse models are based on GALPROP. Appreciation to the
GALPROP collaboration.
Fall 2008 Small suite of files for testing your installation of
EMC2
(otherwise known as Bayes-Image_Analysis_dvd). It contains
sample input data and models, plus logs and outputs of the EMC2-c1 runs on
them.
Get the gzipped tar-file
HERE (BayesEMC2TestSet02.tar.gz).
"Gunzip" the tar files; extract the log files from the tar (and gunzip them);
and look at the beginning of the log-files to see how version "c1" (which required the
input data, model, exposure, and PSF files to be in the form "cnt.txt", "bkg.txt"; etc) were run.
NOTICE that these runs don't actually use the PSF deconvolution property -- just
the multi-scale and fit-testing capabilites.
It contains:
- BayesEMC2TestSet02/do_emc2c1_ModelDataPaddedIntTime539.0.060529_on_Nu3SumPoisDatonPaddedIntTime133.0.070905.it4000.ms100gamttlcnt_1_.05FitS0.070905d.log.gz
- BayesEMC2TestSet02/do_emc2c1_ModelDataPaddedIntTime539.0.060529_on_Nu3SumPoisDatonPaddedIntTime133.0.070905.it4000.ms100gamttlcnt_1_.05FitS0.070905d.output.txt.gz
- BayesEMC2TestSet02/do_emc2c1_ModelDataPaddedIntTime539.0.060529_on_PoisDatonPaddedIntTime539.X.060612.it5000.ms100gamttlcnt_1_.05FitS0.070906a.log.gz
- BayesEMC2TestSet02/do_emc2c1_ModelDataPaddedIntTime539.0.060529_on_PoisDatonPaddedIntTime539.X.060612.it5000.ms100gamttlcnt_1_.05FitS0.070906a.output.txt.gz
- BayesEMC2TestSet02/ModelDataPaddedIntTime539.0.060529.txt.gz
- BayesEMC2TestSet02/Nu3SumPoisDatonPaddedIntTime133.0.070905.txt.gz
- BayesEMC2TestSet02/PoisDatonPaddedIntTime539.X.060612.txt.gz
Please let me know if you have any questions: aconnors AT eurekabayes.com
PREVIOUS TEST FILES:
- Bayes-Image-Analysis_dvd_060818: This contains data runs. See AAA-READ.ME:
- Bayes-Image-Analysis_dvd_060818/AAA-READ.ME
- These first are binned 64x64. These use real CGRO/EGRET All-Sky data. Kind of fun to see the point-sources popping out!
- These use simulated CGRO/EGRET all-sky "datons". These emulate
a faint model-mismatch, with the very broad Inverse-Compton
component suppressed ("squished") ABOVE the Galactic Plane (i.e. "half"):
Note that these were the most difficult cases we tried!!
- These use simulated CGRO/EGRET all-sky "datons". These emulate
a faint model-mismatch, with the very broad Inverse-Compton
component suppressed ("squished") above and below the Galactic Plane:
- These use simulated CGRO/EGRET all-sky "datons". These emulate
BRIGHT discontinuous Pion+Bremmstrahlung gaseous components at right
angles to the Galactic plane. These were the easiest cases we tried.
- These use simulated CGRO/EGRET all-sky "datons". These emulate a
FAINT discontinuous Pion+Bremmstrahlung gaseous components at right
angles to the Galactic plane. This was also a tough case.
*** SPECIAL
NOTE July 18 2007 : All runs using the data-files
"Nu2Sum73PoisDatonPaddedIntTime539.0.060529" were TAKEN OUT, as
one of the data files from which it originated was
corrupted. (Essentially, it was multiplied then cast to integers from
a fits file at an incorrect place in a python procedure, before
"BSCALE" was properly applied.) ***
-
These use simulated CGRO/EGRET all-sky "datons". These emulate
a model that completely describes the data. That is, these datons
were generated from the model used as "baseline".
SO these runs show us what "nothing" looks like.
- Alpha test-code for EMC2. Very untidy! See AAA-READ.ME