# Difference between revisions of "GlueX PID Meeting, December 6, 2018"

(→Agenda) |
|||

Line 18: | Line 18: | ||

==Minutes== | ==Minutes== | ||

+ | Sean gave a talk about how PID was done in CLEO, which has similar variables to gluex. Which lead into a second presentation by Sean on what is currently in the code. Most of the variables reside in DChargedTrackHypothesis however these hooks are in need of correct errors/hit resolutions from various detectors. | ||

+ | |||

+ | Next Beni posted a paper describing how Hermes did PID. Hermes used Bayesian probabilities which utilized "flux factors" as a function of theta and momentum. Peter: need photon energy too. | ||

+ | |||

+ | Next Daniel gave a talk about his multivariate work. He discussed Machine Learning for PID as well as introduced a python framework which uses apache spark+root. The current state does not use best practices but does have many useful diagnostic tools. | ||

+ | |||

+ | Discussing where we go from here we have decided to attempt to vet both PID FOM and Daniel's methods on a couple of channels; Ks->K pi can provide pure pions. Rho should be looked at and Jpsi->ee was suggested as a "null hypothesis for pi/e identification. | ||

+ | |||

+ | Additionally, the detector subgroups will be talked to about providing the necessary variables (tracking meeting) |

## Revision as of 11:19, 18 December 2018

# Meeting Connections

- To join via a Web Browser, go to the page [1] https://bluejeans.com/198066682.
- To join via Polycom room system go to the IP Address: 199.48.152.152 (bjn.vc) and enter the meeting ID: 198066682.
- To join via phone, use one of the following numbers and the Conference ID: 198066682.
- US or Canada: +1 408 740 7256 or
- US or Canada: +1 888 240 2560

## Agenda

- Announcements
- PID @ CLEO (Sean)
- Current PID FOM Status (Sean)
- Multivariate work (Daniel)
- File:Pid hermes.pdf
- Discussion
- AOB

## Minutes

Sean gave a talk about how PID was done in CLEO, which has similar variables to gluex. Which lead into a second presentation by Sean on what is currently in the code. Most of the variables reside in DChargedTrackHypothesis however these hooks are in need of correct errors/hit resolutions from various detectors.

Next Beni posted a paper describing how Hermes did PID. Hermes used Bayesian probabilities which utilized "flux factors" as a function of theta and momentum. Peter: need photon energy too.

Next Daniel gave a talk about his multivariate work. He discussed Machine Learning for PID as well as introduced a python framework which uses apache spark+root. The current state does not use best practices but does have many useful diagnostic tools.

Discussing where we go from here we have decided to attempt to vet both PID FOM and Daniel's methods on a couple of channels; Ks->K pi can provide pure pions. Rho should be looked at and Jpsi->ee was suggested as a "null hypothesis for pi/e identification.

Additionally, the detector subgroups will be talked to about providing the necessary variables (tracking meeting)