2020_01_17 - fermilab-accelerator-ai/meetings GitHub Wiki

meeting summary

• data transfer to PNNL
  • still having trouble with FNAL -> PNNL site to site data transfer
  • currently Jason is scping files to his laptop and loading them to a Google Drive - this allows Malachi and Lucy to operate on some files with all of the data input variables
  • Jason is communicating with Shreyas (FNAL grid team) about how to make progress - collecting information about DTN mount point, etc. so PNNL can make appropriate security allowances for FNAL comms
• tests with the current system
  • Lucy has built and trained a model that suggests some new parameters for the current (autoregressive) system and possibly 10x improvement in errors - we will test these new values on live hardware
• new board deployment discussion
  • AD team is having difficulty securing a test board for full shakedown of the ADCs
  • schedule impact is minimized because PCB layout work is continuing under the assumption the ADCs will be okay (low enough cross talk, etc.)
    • small risk we need to re-do layout and fix the ADCs if this turns out to be false
  • in the meantime, we have a devkit for the ADCs, so we can work on the associated code
• ML on FPGAs
  • ongoing saga with licenses for Intel toolkit (compiler piece seems to be okay but validation simulation component, "modelsim", does not have a powerful enough license)
  • some potential to use Mentor Catapult HLS tools targeting Intel hardware (so, 3rd party solution) - looks promising so far
• Data quality monitoring
  • Jason observed some slow drifts (up to 5 cumulative seconds) in the data, likely connected to line frequency in timing
  • some variables show larger (correlated across variables) jumps with some vague periodic likelihood