Choose timezone
Your profile timezone:
ProtoDUNE pi0 ML
→
America/Los_Angeles
Description
https://stanford.zoom.us/j/6509264799
2019-06-11
===============================================================================
https://indico.slac.stanford.edu/event/145/
Pi0 MC Studies (Hannah)
-------------------------------------------------------------------------------
Pi0 mass reconstruction with g4 truth information.
- why the _truth_ mass w/o SCE not a delta function? (need to check)
- Kazu suggested to use the starting points of the two photons and
pi0 vertex to calculate the mass
Plans for ML Workshop at SLAC
-------------------------------------------------------------------------------
The goal is to **work together** to produce the 1st iteration of full reco chain for pi0.
Kazu will follow up computer accounts for non-SLAC attendants.
We will start with a small pi+ samples for code development.
We should also outline a plan for energy calibration for space points.
Meanwhile we could use SimEnergyDeposit during the workshop.
**Mon 6/17:** - All day, parallel
1. Beginners: General ML lectures for beginners
2. Experts: Working / data preparation
**Tue 6/18**
Morning : Tutorials for ML codes (w/ focus on pi0 analysis)
Afternoon (parallel):
1. code stitching of ghost pts removal + UResNet + PPN, defining output dataformat (Patrick, ...)
2. prepare a _simple_ (cone-based) pi0 reconstruction algorithm using MCTruth
3. rough calibration w/ single MC e-, p, mu particles
4. develop GNN & clustering algorithm (Brad, ...)
**Wed 6/19** - All day
1. integrate ML output to cone-based pi0 reco algorithm
2. develop/integrate GNN
**Thu 6/20** - All day
1. produce performance plots, e.g. pi0 mass, opening angle, ...
2. develop/integrate GNN
3. plan on data calibration
4. review
**Fri 6/21** - All/half day
Overflow, discussion, packing, drinking ...
There are minutes attached to this event.
Show them.
