Please add development news to this page. Probably adding news items as Comments will be most convenient. This is kind of a Field of Dreams approach to gathering information.
Confluence will be down Monday 6AM PT to 7AM PT. We will be adding new login options.
Please add development news to this page. Probably adding news items as Comments will be most convenient. This is kind of a Field of Dreams approach to gathering information.
441 Comments
Toby Burnett
map_tools::gtexpcube
This is a tool, implemented by the code in map_tools/src/gt_expcube/exposure_map.cxx, that creates an "exposure cube", a multilayer image from the combination of an effective area and the livetime cube.
It was not updated when phi dependence was added to both the effective area and the livetime cube implementations.
It turned out that the effective area in the irf P6_V8 differs from P6_V3 only in the addition of a table with the phi modulation, and that the previous version evaluated the effective area only at phi=0. This resulted in a difference in the exposure calculation of 10%. Looking at the Etendue plot for front and back at 1 GeV,
note that evaluating at phi=0 can indeed introduce a 10% difference, which is dominated by back photons. Note that the phi-dependent effective area equals the mean at phi between 15 and 20 degrees for all energies.
So I've updated the code to average the effective area over phi if there is no phi information in the livetime cube, or to do the combined integral over phi and theta if there is. In spite of a lot of effort, I can't get the v3/v8 to be less than around 1% different.
7 Apr 2010. map_tools-07-05-02
Added two hidden parameters:
name
default
description
deltaphi
5
phi increment when averaging over the phi dependence in effective area
limits are 1, 45: for the latter evaluate only at phi=15
ignorephi
no
set to "yes" to ignore phi information if present in the livetime cube
James Chiang
Unknown User (hirayama)
In the pulsar tools development, the following improvement and bug fixes were done since the last report.
James Chiang
Fixed a bug in the BinnedLikelihood calculation noted by Luigi Tibaldo. This had to do with integrating the model counts per pixel. For Luigi's example, this resulted in a disparity at the 1e-5 level for the calculation of the log-likelihood.
Likelihood-16-04-03.
James Chiang
Jean requested a new feature in the binned likelihood implementation whereby model map information for sources that have only fixed parameters be summed and stored as a single entity rather than have the corresponding maps held separately in memory. For example, if 10 sources in the xml model are entirely fixed, then in the aggregate they should take up roughly the same memory as single source. The code will modify the summed model if one of the sources is subsequently set free (in the python interface) and should behave as before, memory usage aside. Currently, this is a one-way operation: fixing a source that was previously free will not reduce the memory usage. This will be implemented next.
Likelihood v16r5p0.
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
To compute diffuse response quantities only for diffuse class events under the pre-P7V3 EVENT_CLASS formatting, one would do
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
Johann Cohen-Tanugi
Bug fix to DMFIT :
DMFIT uses a table derived from DarkSusy. While DS stops at 10GeV in DM mass, DMFIT can go below using a rescaling trick that works until threshold effects become too important.
There was a bug in the implementation of this trick. Now, comparison of integrated yields with Pythia runs shows agreement to better than 1% for the tau channel down to ~5GeV. Below 5GeV, threshold
effects seem to enter into play and discrepancy rises to 12% at 2GeV.
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
Johann Cohen-Tanugi
Several patch tags to deal with gcc 4.4 builds:
xmlBase-05-05-10 sourceIdentify-02-03-02 astro-03-11-09 GRB-04-04-06 GRBtemplate-01-03-04 orbitSim-02-00-07 healpix-02-03-05 map_tools-07-05-06
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
Unknown User (geargo)
Dear Chiang,
I have a naif question. Regarding priors implementation on 09-22-00, I do not sure to understand the following: If I perform a fit with many sources, may I put ?1D prior on each source or just one prior by fit.
Thanks,
German
James Chiang
James Chiang
James Chiang
James Chiang
Unknown User (hirayama)
timeSystem-06-05-00
pulsarDb-09-00-00
pulsePhase-09-00-00
periodSearch-10-08-06
See https://www-glast.stanford.edu/protected/mail/scisoft/0646.html for more details.
James Chiang
James Chiang
James Chiang
Unknown User (lande)
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
Unknown User (hirayama)
Three web pages are added to the pulsar tools development web site to cover the new topics introduced in the latest pulsar tools packages (timeSystem-06-05-00, pulsarDb-09-00-00, pulsePhase-09-00-00, and periodSearch-10-08-06). Eventually those web pages should be a part of the FSSC official web site once the pulsar tools packages listed above are officially released, in my opinion, but for now they are available as below.
Explains the new parameters srcposition and strict of gtophase, in the style of the pulsar tools tutorials. Once the new gtophase is officially released, it is suggested to append the contents to the tutorial for gtophase, Binary Orbital Phase Calculation.
Describes updates and additions to Pulsar Tools Anatomy to cover the changes made in the above listed versions. Once the new pulsar tools are officially released, it is suggested to follow the instructions on the addendum to update the current Pulsar Tools Anatomy.
Explains the D4 file format (FITS and text) in detail. Once the pulsar tools with new ephemeris support are officially released, it is suggested to add a new page under Library >> User Support, right next to Pulsar Tools Anatomy.
James Chiang
James Chiang
James Chiang
Unknown User (hirayama)
The following pulsar tools packages were tagged, and the Release Manager tested them all OK (in ScienceTools-LATEST-1-3442 on rh4-32, rh4-64, rh5-32, rh5-64, and snowl-32).
Note that those versions are supposed to work with tip-02-17-01, which resolved JIRA STARCH-16. Any previous versions of tip will cause the pulsar tools unit tests fail.
James Chiang
Unknown User (hirayama)
Promoted to ScienceTools-HEAD-1-898 (by James Peachey):
Promoted to ScienceTools-HEAD-1-899:
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
Here is an example that divides the power-law model into three energy bands. Note that the spatialModels should match.
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
Unknown User (hirayama)
Tagged pulsarDb-09-00-02.
Promoted to ScienceTools-HEAD-1-927:
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
One can enable the map-based diffuse response calculation similarly,
James Chiang
Since the code just checks for the existence of these environment variables and for their values, in order to disable either of these options, one must unset the corresponding environment variable, e.g.,
James Chiang
To avoid the map-based integration, simply omit the attribute:
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
Johann Cohen-Tanugi
-bash-4.1$ stag -notes="First tag including the HEALPIX extension to evtbin" -minor -cvspath=evtbin
Successfully made tag evtbin-02-06-00 along branch MAIN
-bash-4.1$ stag -notes="First tag including the HEALPIX extension to evtbin" -minor -cvspath=evtbin
Successfully made tag evtbin-02-06-00 along branch MAIN
-bash-4.1$ stag -notes="First tag including the HEALPIX extension to evtbin" -minor -cvspath=evtbin
Successfully made tag evtbin-02-06-00 along branch MAIN
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
Johann Cohen-Tanugi
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
Johann Cohen-Tanugi
st_facilities-00-20-02 : copy over irfs/latResponse Bilinear and FitsTable helper classes, so that they are visible in GlastRelease land. These copy are within the namespace st_facilities, while irfs code continue to use the copies in the latResponse namespaces. This will be fixed when the dust settles
Johann Cohen-Tanugi
Likelihood-17-27-00 : the new DMFIT is in. Check https://confluence.slac.stanford.edu/display/SCIGRPS/New+DMFIT
James Chiang
James Chiang
James Chiang
James Chiang
Stephen Fegan
James Chiang
James Chiang
James Chiang
Stephen Fegan
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
Modifications to support automatic IRF discovery
James Chiang
James Chiang
Modifications to support FSSC-requested irfs handling:
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
that define EARTH_AZIMUTH_ANGLE correctly. Source xml definitions for EarthPhenom sources should use
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
The output directory and output files should be given in the user-provided setup file:
Front and back event selections must be specified in the Prune.cuts explicitly. The following would correspond to SOURCE events (in P7[REP]) for the front section:
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
Johann Cohen-Tanugi
Added note : gtexpcube2 now creates a SKYMAP named HEALPIX extension automatically, if the cube in input is in healpix. This means that no new input parameter has been added to gtexpcube2 par file.
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
Added a flag in the setup.py file to use just the front psf scaling for all events:
James Chiang
Implemented an alternative point source map calculation that uses a PSF value that is averaged over the theta-extent covered by the target pixel. To enable, set an environment variable:
To disable again, unsetenv USE_NEW_PSF_ESTIMATOR.
James Chiang
James Chiang
James Chiang
For some reason, this tag does not actually include tip-02-18-05. Will tag a new release once this is understood.
James Chiang
Giacomo Vianello
James Chiang
James Chiang
James Chiang
James Chiang
allow users to set scaling function parameters for PSF and edisp via setup.py and propagate to FITS output via root parameter file.
Example setup.py code:
If the user omits any of the three sequences, PSF.pars, Edisp.front_pars, or Edisp.back_pars, the default values (shown above) are used.
The parameters used are printed to the screen by makeirf:
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
This is set via the apply_edisp attribute, e.g.,
If the apply_edisp attribute is missing, then energy dispersion can be applied. Valid values for apply_edisp are only "true" and "false".
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
matplotlib support from Tom Stephens. For gtlike, this would only be available (currently) from non-ROOT builds. To enable in pyLikelihood, do
James Chiang
James Chiang
James Chiang
James Chiang
Add energy dispersion handling to gtmodel. Enable by setting the edisp hidden parameter:
James Chiang
Users should note that in the recent release, ScienceTools-09-34-03, one can turn on/off energy dispersion handling in gtmodel already using the environment variable:
Note that this environment variable applies to all binned analyses within the current session.
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
Fixed LK-94@JIRA: pyLikelihood does not propagate appy_edisp attributes to the output xml file
James Chiang
James Chiang
Example:
The attribute must be set in the "spectrum" tag. Use of the command line option, edisp=no, turns off energy dispersion for all sources, regardless of the individual apply_edisp flags in the xml model.
James Chiang
James Chiang
enable array values in the merit tuple to be accessed by index. This enable entries in the merit-to-FT1 dictionary of the form
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
Gudlaugur Johannesson
What does that mean in terms of manually setting IRFs in gtexpcube2?
James Chiang
You need to specify the evtype value. Other changes here
James Chiang
James Chiang
add Saclay implementations of angular distance, called saclay_dist
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
Here is an example xml definition:
Constraints:
The Index# and Break# parameters must appear in order in the xml definition.
There must be one fewer Break parameter than Index parameters.
The Break parameters all must have free="0", i.e., they must be fixed.
The Normalization is the function value at the first break energy, Break0.
The maximum number of Index values is 50.
James Chiang
James Chiang
Example xml definition:
James Chiang
James Chiang
Enable user to apply a scaling function to the source spectrum in the xml model definition to enable effective area systematics studies. The scaling function is implemented as a FileFunction object. Example xml model definition:
The scaling_file must be an ascii file with two columns: energy in MeV and scaling factor. The energies must cover the entire range of the analysis or an exception will be thrown. Points are interpolated on a log-log grid, i.e., a power-law is used to interpolate between points.
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
Giacomo Vianello
James Chiang
James Chiang
James Chiang
From Eric Charles (https://www-glast.stanford.edu/protected/mail/infrasoft/2980.html):
Johann Cohen-Tanugi
I presume that 2 packages need to be added to GLAST_EXT, for these to work. Is it possible to add the "official" paths to them here as well?
James Chiang
I'm not sure what you mean by "official" paths. The locations of the various extlibs are in allExternals.scons in the SConsFiles package:
http://glast.stanford.edu/cgi-bin/viewcvs/SConsFiles/allExternals.scons?r1=1.167&r2=1.168
Is that the info you want?
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
Giacomo Vianello
James Chiang
Giacomo Vianello
Giacomo Vianello
James Chiang
James Chiang
Likelihood-19-01-13, pyLikelihood-02-00-06
LK-102@JIRA: request to implement SummedLikelihood::NpredValue
James Chiang
James Chiang
James Chiang
James Chiang
James Chiang
Johann Cohen-Tanugi
James Chiang
James Chiang
Matthew Wood
Implement SpatialDisk and SpatialGaussian models as analytic representations of a 2D disk and gaussian respectively. Example XML definitions given below:
Unknown User (jurgen)
Good to see that disk and Gaussian spatial models are now also available in the Science Tools.
I have these also implement in GammaLib / ctools and would like to have an XML format that is compatible with the Fermi/LAT Science Tools. Here the XML definition that is current used by GammaLib:
<source name="Gaussian Crab" type="ExtendedSource">
The names I invented were inspired from the "SkyDirFunction" used for a point source. Note that I also have elliptical models and they are called "EllipticalDisk" and "EllipticalGauss", so the names "RadialDisk" and "RadialGauss" for models that depend only on radius would also make sense. By the way: what is the source type that is implemented in the Science Tools?
Any possibility to have common names so that XML files can be easily exchanged between Science Tools and ctools?
Matthew Wood
This is certainly possible. I didn't give much thought to the names and in retrospect I probably should have looked at what you were using in ctools. The classes aren't widely used yet (we only made the 11-00-00 release available a few weeks ago) so changing the names shouldn't be too disruptive at this point. However we should probably try to converge on something relatively soon.
The SpatialDisk and SpatialGaussian classes use the existing DiffuseSource type so I guess that will also be an issue for making the XML models compatible. Creating an ExtendedSource type would be a substantial amount of work but maybe a possible solution would be to make ExtendedSource an alias for DiffuseSource.
Unknown User (jurgen)
You can see how GammaLib organises the model classes at http://cta.irap.omp.eu/gammalib/user_manual/modules/model.html
I have three types of models: "PointSource", "ExtendedSource" (used for radial and elliptical models) and "DiffuseSource" (used for maps, cubes, isotropic). The basic idea is that the spatial distribution of an "ExtendedSource" can be parametrised, while that of a "DiffuseSource" can not be parametrised. The good news is that I actually don't use the type in the software, but of course it would be good if the XML files are consistent. If a proxy could work on your side this would be perfect as it preserves the logical structure of the different model types.
Concerning the names of the models, I don't like mine (the "Function" names), but since I have also elliptical models specifying "SpatialGaussian" would not be enough. I think also that the "Radial" names would be more precise ("RadialGauss", "RadialDisk", I also have a "ShellFunction" which could become "RadialShell"). I would certainly implement some fix in the ctools/GammaLib code that supports the old and the new names, but maybe dump a warning that the old names are deprecated.
James Chiang
This includes
James Chiang
Matthew Wood
Change naming convention for 2D gaussian and 2D disk spatial models to agree with model names used in GammaLib/ctools. Example XML definitions given below:
Matthew Wood
James Chiang
Matthew Wood
latResponse-04-02-02
Changes to interpolation scheme in PsfIntegralCache to improve accuracy of npred calculation.
James Chiang
Eric Charles
Eric Charles
Matthew Wood
irfInterface-03-03-00
st_facilities-00-22-00
Add RootFinder utility class that solves for roots of 1-D functions using brent solver from GSL.
irfInterface-03-03-01
Bug fix for angularContainment to ensure unit normalization of PSF integral.
James Chiang
Matthew Wood
Likelihood-20-08-08
James Chiang
Matthew Wood
irfInterface-03-03-02
latResponse-04-03-00
Move header files to expose IRF classes to pyIrfLoader.
Add setter/getter functions for retrieving and updating IRF parameters at runtime.
pyIrfLoader-00-02-00
Expose current IRF implementation classes (Psf3, Edisp3, Aeff) from latResponse.
Add methods to cast abstract interface classes to derived classes from latResponse.
Matthew Wood
latResponse-04-03-01
Added CompositeSource and associated functionality to Likelihood.
Implement new padding scheme for diffuse map convolution.
Change column names in gttscube output file.
pyLikelihood-02-04-00
Removed BinnedLikelihood2.
Add option to construct Likelihood from BinnedConfig object.
James Chiang
James Chiang
Matthew Wood
Likelihood-20-09-03
Fix bug BinnedCountsCache that was preventing construction of weighted counts cube.
Likelihood-20-09-04
Fix typo causing a large vector to be passed by value and slowing Likelihood evaluation down dramatically
Likelihood-20-09-05
James Chiang
Eric Charles
James Chiang
Matthew Wood
Likelihood-20-09-07
Bug fixes for behavior of source map cacheing when dealing with fixed sources or source maps set with setSourceMapImage.
Likelihood-20-09-08
Changes to address const-correctness issues in FSSC builds.
st_facilities-00-22-02
Bug fix to address failures in unbinned analysis. Protect against case that xx == x.back() in Util::interpolate.
James Chiang
Matthew Wood
New spectra model PLSuperExpCutoff2 that replaces the cutoff energy with an ExpFactor parameter (see LK-123). The functional form of this model is
Example XML definition:
Eric Charles
Likelihood-20-12-03 Changes and fixes requested by Jean Ballet:
evtbin-03-00-01 Change keyword to define HealpixRegion to HPX_REG to match Matt's conventions
pyLikelihood-02-04-03 18-Aug-2017 Fixes for LK-126 and LK-127
Matthew Wood
irfs/handoff_response-06-02-00
Added support for ingesting merit files from a file list (skipping the prune step).
Refactored IRF generation scripts to simplify configuration.
Automatically determine number of thrown events from list of input files and jobinfo tuple.
irfs/caldb-07-01-00
Added functionality in build_caldb.py to install IRFs into an existing CALDB directory structure.
James Chiang
Matthew Wood
Eric Charles
Likelihood-20-13-03. Includes the following change since Likelihood-20-12-03 on 15-Sept-2017
James Chiang