Date: Wed, 3 Oct 2001 17:42:30 -0400 (EDT) From: Megan Donahue Subject: SHARE Town Hall Meeting at Noon notes To: gak@stsci.edu Cc: donahue@stsci.edu MIME-Version: 1.0 Content-MD5: Th3WAgzQAcIhQqSRAxuWmQ== SHARE Town hall meeting Noon, Auditorium Gerry Kriss presented the SHARE recommendations and opened the discussion to comments and input. Replies were typically from Gerry. Notes: I paraphrase and summarize discussion items below. Some comments have names associated, as I caught them, but I'm not a transcriptionist. Eric: What about the community resistance to "black boxes"? Reply: Documentation is very important for the community to assess the algorithms [I would add source code with comments is important too.] Torsten: How is the SHARE activity different from pipeline improvement going on in the instrument groups? Reply: SHARE crosses instrument boundaries e.g. instrument combination. Mark: Overlap is OK; SHARE recommendations can be made for pipeline improvement. [Note: Rodger commented later in the meeting that SHARE is not a project in and of itself -- it was a means to collect ideas. Exactly how those ideas are implemented is not restricted. Pipeline work by instrument groups can be affected; community can contribute as well.] Harry: Prioritization is needed; prioritization should take in the big picture view. For example, cleaning up WFPC2 images, CR-cleaning, hot pixel cleaning is important. Was it left out on purpose? There has been some progress in WFPC2 made by ECF. Megan: ECF devised post-hoc WFPC2 associations. Some percentage of these can be used to produce CR-cleaned images; most however are merely associated. This project, taken one step further, would produce CR-cleaned images from individual frames which are NOT perfectly aligned. Bill: We are locked into a single "pipeline" concept. We could have different pipelines for different purposes; or merge pipelines with very similar purposes (e.g. WF3 data could be calibrated in an ACS pipeline). Torsten: One has to be careful (regarding multiple pipelines) that we have one final science product. Rodger: Was there thought about Sloan? (Being here). Mark: We could use Sloan or radio source catalogs to get absolute astrometry (a very high priority SHARE recommendation). Sloan may provide some means to improve the WCS. But it's ground-based photometry... We should also investigate how Sloan did cataloging. Marc: SexTractor would be much easier to implement than the proprietary Sloan algorithm. Torsten: What's the problem with the headers? Reply: Pointing accuracy and distortion. Ken: Items 3-4-5 on your list are more judgement-dependent science activities. What will be the process for deciding how to go beyond the standard processing? How will you convince the community that your way is right? Marc: Such cataloging for example would not be useful or reliable for every scientific purpose (crowded field photometry). It's important to be clear HOW it was done. Mark: We need to go beyond single frame reduction. Catalogs could be useful for every image even if it is to classify the image itself (e.g. sparse, crowded, extended emission, other classifications) for queries. Gerhardt: I'd worry about selection effects. ?: What about line IDs in spectra? Discussion: how to improve coordinates from Guide Stars. (Some implementation schemes discussed.) Harry: Some of the SHARE items might be applicable only to a subset of the data. Harry: Would some of these SHARE recommendations be amenable to human effort - hiring someone to take on the task, human quality control? In my experience, cataloging requires human inspection. Fully automated cataloging is not very useful. Rodger: SHARE is not a project ...(see above) Danny: Were any DAs or Archive Specialists involved? Rodger: Science input is required up front, what the scientists need to do their work. Implementation may involve more technical experts. Rodger: We should think about how to get community results back into the pipeline, into the archive. Harry: SeXtractor is moving to open source. Gerhardt: so is focas (which is being improved too.) Mauro: A comment regarding implementation - we should separate the general features (capabilities?) of the pipeline from instrument-specific ones. Gerhardt: We could request calibration or archive research proposals internally to improve the calibration process. Danny: Reduction issues seem to be more popular than analysis. Marc: But look at Sloan -- the catalog is very popular and nobody's lining up to re-create it from the raw data. 1-2 images may have special applications. Rodger: We are starting these Treasury programs (observing and archives) Mark: X-ray catalogs are routinely derived from individual X-ray observations. The observers get them but they don't publish them. They're first cut. A first-cut "non-scientific" catalog is useful for registrations, astrometry, mosaicing. The next level could categorize an image as crowded or sparse, which is a useful queriable property of the image. Ken: Make sure the user understands the actual value of the catalog. Discussion: catalogs, pros/cons, community use, is it useful? to whom? for how much of the data? Harry: What is the priority setting process -- what should it be? Reply: SHARE internally prioritized its recommendations based on perceived scientific usefulness of the data product. But it did not address what the overall prioritization process should be? Perhaps there could be internal proposals to do specific projects?