ADASS Day #1

These are my opinions and thoughts about the first day of ADASS.   It isn’t necessarily a summary of the day or complete but  just my reflections on the day along with some random commentary or thoughts.  I might not do this for other days.  I might.  Also these thoughts might not be in chronological order and can also be pretty random.  

1. The day definitely started out very interestingly with an opening by a local elder followed by a talk on Machine Learning and robots by Hugh Durrant-Whyte.   The Acknowledgement of Country was poignant and funny.   It was followed by a talk that was initial focused on how to use machine learning to extract minerals from the ground.   It’s only tonight I realize how jarring a transition it was.   Neither of these was about astronomy per se, but both seemed to reflect issues that the astronomy is facing (de-colonialization and the need to  commercialize skills).

2.  I’ll group talks by Alessandra Aloisi, Lisa Storrie-Lombardi, and Jesus Salgado together as they were all great talks about various archives.  The amount of data ‘freely’ available to anyone is amazing and the growth in the archives is impressive.  As Aloise pointed out, 60% of HST papers are coming from the archive.   So much science is enabled by open access to the data.   I’ve used at least the MAST and Spitzer archives  for my research (among many others) and for my students’ research.   And the demo by Bruno Merin of ESO Sky was just cool.

3.  I was left wanting so much more out of the talks.    So much more details and questions that went un-answered and I’m hoping to meet and have conversations with Guido Cupani about Espresso (wavelength calibration and extraction?  How are things done differently for high precision),  Mario Jaric about LSST software (design and development and packaging), William O’Mullane about Docker (plus/minus vs. other methods), and Tamas Budavari (how can streaming analysis be applied to other data sets?).  Actually that last question might not be so much for Budavari, but for his student, Matthias Lee, whom he brought on in the middle of the talk.    I’m actually super interested in that as it relates directly to another project I am currently not working on at the moment.

4.   I spent a good part of my day collection gender information on speakers and people who ask questions for James Davenports gender study on conference talks.   Overall, there were 11 male speakers and 2 female speakers and 28 male questioners and 4 female questioners.   Those numbers are a little thrown off as Tsianheng Liang (Extracting filaments) and Jessica Mink (BoF: data formats) were not able to attend.

5.  Some good overall challenges were issued today which I’m writing down to remember them for later (also probably missing many of them):

  • Josh Peek:  how do we hypothesis *generate* with large unwieldy data sets?
  • Arna Karnick: how do we reward contributions to open source projects?
  • Nuria Lorente: How do we increase the gender repsentation? (And I’ll add the representation from all groups?)

6.  I did attend the data formats Birds of a Feather (BoF) session.   I’m really happy there are other people thinking about standards as this is a pretty thankless but a necessarily job.  But I am also happy to working in the astropy environment where people have created some great tools for reading in data and working on different data models.   The astropy.table can import almost any format (ascii, latex, csv, FITS, HDF5, those machine readable tables that come with papers by just giving it the README–I really love this feature, and so many more) which has made life much easier.  Plus making tables is easy and then can be written out in any format you or your colleagues want.   In the meantime,  for things like the ccdproc package we haven’t focused on how to read in FITS files at every stage, but read it into a CCDData object (based on astropy’s NDData object) and then do all our operations on that or a numpy array.  The code is fairly agnostic even to the data model without having to do too much replication (of course it has to be a numpy-array like object).

7.  There were some mutterings about programming languages during two good talks on detecting diffuse objects by Mohammad Akhlaghi (NoiseChisel) and Tony Butler-Yeoman (Oddity).    I obviously have my favorites and am not above throwing out a comment about one language or operating system or data format or editor but nowadays, languages are so interchangeable (see Juypiter notebooks for example) and wrap-able that you find the right one for you and for the job that needs to get done and it doesn’t matter.  Plus it is fun learning new languages.  But you will pry vi from my cold, dead fingers clutching ‘:wq’.

8.  And one final talk that particularly resonated with me as a pipeline writer for a telescope that hasn’t always worked, Christian Wolf talking about Skymapper.   I’ve faced many of those problems myself as well or similar things with a telescope that took longer than expected to be fully operation and an under-resourced pipeline.   I think Josh Peek summed it up best though:

But it was great to see that the telescope and pipeline were both now operation and working well.

Okay off to bed to wake up in time to listen to tomorrow’s first talk.  I haven’t even gotten to the posters yet or the conversations (in person or on twitter).