This topic thread is for meeting summaries of the “Photo-z Commissioning Team”. Members of this team are working with the Rubin Observatory commissioning effort under an MOU (see SITCOMTN-010 for more information).
As described in the Forum topic “Rubin Commissioning and the Photo-z Validation Cooperative”, this is the team that will take the lead on guiding implementation and performing early science validation for the shortlisted PZ estimators with the unreleased commissioning data, which will help the science community make use of Data Previews 1 and 2 during the Photo-z Validation Cooperative phase of the “Roadmap to Photo-z for the LSST Object Catalog”.
The team wants to keep any interested parties in the broader science community informed of their ongoing work, so updates will periodically be posted here in the Rubin Community Forum.
Tue Nov 8, 2022
Attending: Melissa Graham, Markus Rau, John Franklin Crenshaw, Julia Gschwend, Nacho Sevilla, Sam Schmidt, Alex Malz, Bryce Kalmbach, Shahab Joudaki, Eric Charles, Drew Oldag
Round table of introductions
Review the PZ Commissioning Team MOU
Timeline for commissioning and data previews
Review of the PZ Roadmap and Letters of Recommendation
a. Discussed potentially adding FlexZBoost, for which an LoR was not submitted.
b. Discussed how the team may need to prioritize RAIL-compatible codes.
PZ Server, Julia Gschwend
a. Currently in development at LineA, it is a community resource for photo-z validation.
b. The team is invited to have user accounts and test the tutorials.
Tue Nov 29
Attending: Julia Gschwend, Nacho Sevilla, Alex Malz, Drew Oldag, Melissa DeLucchi, Anja von der Linden, Sam Schmidt, Bryce Kalmbach, John Franklin Crenshaw, Eric Gawiser, Melissa Graham
SIT-COM Science Units: The PZ Commissioning team is part of the SIT-COM Science Unit “Galaxy photometry and photometric redshifts”, co-led by Dan Taranu and Melissa Graham.
Data Preview 0: The first steps for PZ Commissioning team will be to test use of RAIL and the PZ Server from the Rubin Science Platform (RSP) deployed at the Interim Data Facility (IDF; Google Cloud), with the DP0.2 data set. To get started, the relevant DP0.2 tutorial notebooks are numbers 1, 2, and 4a/b.
Initial action items were volunteered by a few team members and the next meeting will be a longer (2h) co-working session in Jan 2023.
Wed Jan 18
Attending: Melissa Graham, Drew Oldag, Julia Gschwend, John Franklin Crenshaw, Alex Malz, Sam Schmidt, Shahab Joudaki, Brant Robertson, Markus Rau
Mock training sets and PZ Server: A new notebook has been added to the photoz folder in the delegate-contributions-dp02 repo demonstrating how to create a mock training set from DP0.2 data, and then made an initial attempt to upload this to the PZ Server. We discussed the process and options for such uploads to the PZ Server.
Running PZ estimators at scale: Options for running scalability tests, evaluating outcomes, and potential problems we might encounter were discussed. Current implementations with RAIL and qp, and how LINCC might best contribute, were also discussed. General agreement on working together with experimentation and benchmarking.
Markus, Alex, and Drew and Sam (and perhaps others) were going to tag up later to discuss practical places to start as well as scope of testing.
Tue Apr 11
Attending: Melissa Graham, Sam Schmidt, Drew Oldag, Nacho Sevilla, Melissa Delucchi, Olivia Lynn, Julia Gschwend, Brant Robertson, John Franklin Crenshaw, Alex Malz, Bryce Kalmbach
Potential 2023 LSST Photo-z Virtual Symposia: we might host a two half-day sessions (e.g., June/November) with a focus on promoting software interoperability for the various LSST PZ-related efforts ongoing within in-kind contributions and Science Collaborations. Melissa G. and Eric C. will take the lead on organizing. Stay tuned for an announcement.
PZ Server & RAIL: Julia G. reported on updates to the LIneA PZ Server related to user uploads, libraries, and tutorials. Updates to the training set maker are currently under construction. Julia G. also reported on progress with ingesting the DP0.2 Object catalog and RAIL runs (FlexZBoost estimate) at LIneA’s clusters. Bug fixes and run-time scalability were discussed. Several action items for co-work between LIneA, RAIL, and LINCC were identified.