Present

Giuseppe Cerati, Vito Di Benedetto, Lynn Garren, Thomas Junk, Guilherme Lima, Gianluca Petrillo, Erica Snider, Hans Wenzel

Remote

Erin Conley, Katherine Lato, Tracy Usher

Hall of infamy
(“sorry I'm late”)

Lorena Escudero, Robert Hatcher, Robert Kutchske, Gleb Sinev

Project status report [Erica Snider]

  • minutes of these meetings are collected in https://larsoft.org
  • "new" version of Geant4 is available: are we interested?
    • [Hans Wenzel] 4.10.4 would allow to select physics lists (see release notes!)
  • script contributed by Robert Hatcher to solve linker errors of "undefined symbol"
    • [Thomas Junk] I wrote a silimar script and ended up not using it because I would miss the translation from the library path into the directive to be actually written in the CMakeLists.txt

Conclusions

  • request for comments on the proposal of upgrading to Geant4 10.4
  • will add directions on how to translate a library path into a library name for CMakeLists.txt

Using proxies to access associated data [Giuseppe Cerati]

  • proxies are helper classes designed to facilitate access to connected data products
  • [Erica Snider] why proxy of proxies is not supported? [A] it is a limitation of the current design, that the author is willing to revisit once C++17 standard is available for LArSoft
  • [Robert Kutschke] is the association proxy using Saba Serish tools? [A] no, because a different type of view is needed
  • recommending to use proxies for tracks, which will make the code more stable with respect to the planned track interface changes
  • [Robert Kutschke] could proxy satisfy the needs for description of multiple interactions? [A] after some discussion, it seems LArSoft is "one data product away" from having a solution; in the current state, there is no facility to describe that

SVM-based track/shower identification in Pandora [Lorena Escudero]

  • discrimination of tracks and showers was added in June 2017, based on a machine learning technique
  • updated to classify a particle flow element (3D) rather than the single clusters
  • [Gianluca Petrillo] what does an experiment need to train the discriminant? [A] there is an option for that, and lot of documentation in Pandora site (in GitHub)
  • [Gianluca Petrillo] would adding a reliable charge from induction planes improve the performance? [A] the performance is quite good already; an improvement from induction planes is not excluded but a test is needed to actually know

Conclusions

  • no objection against the integration of the feature branch