Source detection and catalog pipeline architecture

1

• Source detection and catalog pipeline
  architecture
• Relative merits of the algorithms under
  consideration
• Review of selection criteria and time scale for
  selection
• '3D' methods (map energy, map time)
• Inclusion of a 'source finder' tool into the
  Science Analysis Environment ?
• Pipeline script, interfaces between the
  successive stages
• Operational issues (scheduler, database,
  computational requirements)

2
Relative merits of the algorithms under
consideration

• Several algorithms have been proposed (likehood,
  several Wavelet methods, Bayesian blocks, ). We
  need to have a good understanding on the quality
  of each of these methods.
• Some of them are not included in the pipeline and
  some works need to be done if want to use them.
• The optimal filter has the advantage to produce a
  significance map.
• Such a map could also be derived from the wavelet
  coefficients but in relatively heuristic way.
• It is clear that whatever the selected method for
  source detection, a maximum likehood should be
  used to estimate the other parameters
  (photometry, ).
• It may happen that several algorithms could be
  useful. For instance,
• for a quick look analysis, we need a fast
  method, but not necessary optimal.

3
Review of selection criteria and time scale for
selection
For the selection of the detection method, we
agreed that it would be useful to first make a
set of tests on toy models and make an evaluation
of the different algorithm. Then we can decide
which algorithms should be included in the
pipeline and tested in a more realistic way.
Two data sets already exist. A third one is
needed, which would allows us to see the resolved
power of each method.
4
(No Transcript)
5

1. '3D' methods (map energy, map time)

We have discussed the need of 3D tools (spatial
time or spatial energies) and 4D tools (spatial
time energy). It seems that 4D analysis is
not needed. For temporal analysis, we can
consider a given energy band and make the
temporal analysis for this band only. Several
tools/idea have been discussed. The Voronoi
methods could be used, but we have to figure out
how to use the output of the Voronoi method. Gino
has a project of developping a 3D mexican hat.
I have reported some ongoing effort to develop a
3D wavelet analysis for GLAST.
6
(No Transcript)
7
z 0
z 29
8

• Inclusion of a 'source finder' tool into the
  Science Analysis Environment ? NO
• Pipeline script, interfaces between the
  successive stages.
• The best is to use the GLAST catalog format.
• Operational issues (scheduler, database,
  computational requirements)
• The likehood method is not fast enough for
  finding the source positions but Jim has some
  idea to spead up the algorithm.
• There is a need to build a tool which create
  a diffuse emission model of a part of the sky.