Density estimation of missing data using OT

Malte Algren Geneva, On ATLAS

Interesting.

Notes

  • Malte Algren
  • "Density estimation of missing data using OT"
  • PhD student in Geneva HEP
  • Related to "In painting" problem in
  • Setup the ABCD problem
  • Crop out 4b SR replace removed events with 3b SR
  • Fit OT from 3b -> 4b… Iterate
  • How many steps ? …. How long for each step ? …
    • What do you technical get in the end ? OT(map)
  • How do you deal with point clouds ? … Yes, they use point clouds.
    • 16D → 7D
  • How do you predict another variable ?
  • Suggest adding 8th SvB variable and then comparing that prediction to the rewieghting
  • Very Sensitive to the CR boundry
  • Maybe start with a better infill assumption … eg: the reweighted prediction
  • Application generate synthetic data in the SR

Follow-ups

Links:

202504121742