BlueJeans: https://bluejeans.com/426716450
attending: Eric Bellm Ian Sullivan Unknown User (cmorrison) Meredith RawlsUnknown User (gkovacs)John ParejkoJohn SwinbankUnknown User (emorganson) Krzysztof Findeisen
Topics for discussion
- HiTS reprocessing (Meredith Rawls):
- poking at the recent re-run of ap-pipe, continuing to dig into the photometric calibration. trying to exclude saturated sources. looking at calibration factor (x instFlux = flux); within nights there is variation that seems to have the appropriate shape for rising and setting fields
- PhotoCalTask is essentially just subtracting the reference mags from the inst mags with a little cleanup
- John Parejko wonders if we have a non-photometric night that we're not handling effectively
- Eric would like us to compare our calibrated fluxes for the matched reference sources to the refcats by source color, airmass, seeing. Seems like an obvious check. Does Lauren have a plot ready-made? JDS also suggests looking at what the debug code in PhotoCalTask is doing.
- we discovered last week that obs_decam has no color terms.
- JDS wonders whether this is likely to have a large effect given the small range of airmasses.
- Unknown User (emorganson) wonders if some of the DA WD calibration might have other color terms
- DES processing (Unknown User (emorganson)):
- got Maria's stuff working on Kubernetes, doing other Kafka stuff for EFD
ap_
association (Unknown User (cmorrison)):- working on getting dax_ppdb to output pandas. Some by-hand queries may need some special attention
- config to provide a switch between afw tables and pandas interfaces
- Metrics & ap_verify (Krzysztof Findeisen ):
- image differencing (Unknown User (gkovacs) ):
- returning to image differencing after a tangent on a ticket
- talked to Colin, who reminds us that KBMOD folks are running the Lori Allen dataset (so having us reproduce his work ourselves is not necessary)
- John to ping Yusra for a chat
- DCR (Ian Sullivan ):
- vacation last week, but some progress in the week before:
- changed an internal normalization factor that sped convergence notably–has a ticket to understand (DM-18742)
- shows dependence on false positive rate on range of input seeing
- Review CI (https://chronograf-demo.lsst.codes/):
- AOB
- Eric: HiTS labelled set has inconsistent internal IDs between the features and the objects, so this is stalled. Could just crossmatch it himself, but starting to get to diminishing returns...
- Eric also started thinking about DIASource counts in the style of DMTN-006 etc.; will make a ticket for a notebook, and then think about how to fold into metrics