Agenda And Minutes for 3/17 10am CDT¶
Note that we're now in Daylight Saving Time in the USA!
1. Dial Toll-Free Number: 866-740-1260 (U.S. & Canada)
2. International participants dial:
Toll Number: 303-248-0285
Or International Toll-Free Number: http://www.readytalk.com/intl
3. Enter your 7-digit access code, followed by “#”
Your access code: 7691719
- Photo-z Updates
- Computing Completeness
- Mask Updates (if available)
- SG separation challenge round 2 results
Attending: Ricardo, Aurelio, Robert G, Ramon, Anne, David, Flavia, Diego, Alex D-W et al.
(Eli was out, so we took advantage to go to recess 20 minutes sooner...)
Star-Galaxy separation round 2:
Results are in and we have a bunch of machine learning methods (TPZ, SkyNet, Boosted Decision Trees) which are a real improvement with respect to the DESDM codes. However SPREAD_MODEL and MODEST (in DESDM) do not need training, which means that, as long as PSF is well determined, they'll have the advantage of not needing truth fields which might not be representative. Round 3 will be done in the mid-term, either adding a QSO master list to the current sets OR using stripe 82 and Y1P1 data (though in this case we do not understand all the details of what is coming out yet).
Detection Completeness on SVA1:
Discussion on how this has been done in the past, in an object-by-object basis with DC6, CFHTLS, [[des-lss:Systematics_tests#Detection-completeness|COSMOS]]. Should move on to draw a 2D magnitude-surface brightness completeness plot to take into account also the detectability of low surface brightness objects (Diego).
Summarizing the outcome of the clusters-GE meeting last week: how can we evaluate detection completeness?
1 - Build a 10-tiling version of a DES deep field (SN, COSMOS). Has the advantage of using same photometric system BUT requires a lot of manpower, not good to understand artifacts, influence of different exposure times.
2 - Simulate images and run through current DESDM pipeline. Can change different parameters as we like BUT requires a lot of manpower, are we sure of simulating realistic image features?
3 - Use a deeper field (CFHTLS Deep). Less work BUT needs to translate photometric systems, need to be careful with astrometric differences too.