for DUNE FD 28 th October 2019 Mousam Rai Supervisor: Prof John - - PowerPoint PPT Presentation

β–Ά
for dune fd
SMART_READER_LITE
LIVE PREVIEW

for DUNE FD 28 th October 2019 Mousam Rai Supervisor: Prof John - - PowerPoint PPT Presentation

Jul 08, 2023 153 likes β€’578 views

Track/Shower Discrimination Refresh in Pandora for DUNE FD 28 th October 2019 Mousam Rai Supervisor: Prof John Marshall 1 Roadmap for this presentation The Problem MicroBooNE Variables Current Approach In Pandora BDT1, BDT2, and

slide-1
SLIDE 1

Track/Shower Discrimination Refresh in Pandora for DUNE FD

28th October 2019

Mousam Rai Supervisor: Prof John Marshall

1

slide-2
SLIDE 2

Roadmap for this presentation

  • The Problem
  • MicroBooNE Variables
  • Current Approach In Pandora
  • BDT1, BDT2, and BDT3
  • The End.

2

slide-3
SLIDE 3

The Problem

Track vs Shower discrimination

3

slide-4
SLIDE 4

Track-like (πœˆβˆ’) Shower-like (Ξ³) Argon nucleus fragments from deep inelastic scattering Track-like (π‘ž) Ξ³, p, πœŒβˆ’, Ar- fragments merged as

  • ne

Multiple tracks merged into

  • ne

U View V View W View

4

slide-5
SLIDE 5

MicroBooNE Variables

What are they?

5

slide-6
SLIDE 6

Types of MicroBooNE Variables

  • 8 Topological Variables:
  • length
  • diff
  • gap
  • rms
  • vertexDistance
  • diffAngle
  • pca1
  • pca2
  • 2 Calorimetric Variables:
  • charge1
  • charge2

6

slide-7
SLIDE 7

length (cm) diff (N/A) length – 3D length of the PFO diff – Mean difference between the position of the hits and a straight line, divided by the straight line length

7

slide-8
SLIDE 8

rms (N/A) gap (N/A) gap – Average max gap distance, divided by straight line length rms – Average root mean square of linear sliding fit, divided by straight line length

8

slide-9
SLIDE 9

vertexDistance (cm) diffAngle (rad) vertexDistance – Distance between the PFO vertex and the primary vertex diffAngle – Difference between the opening and closing angles calculated over 50% of the pfo closest and furthest from the vertex.

9

slide-10
SLIDE 10

pca2 (N/A) pca1 (N/A) pca1 – Ratio between the second largest and the largest PCA eigenvalue pca2 – Ratio between the third largest and the largest PCA eigenvalue

10

slide-11
SLIDE 11

charge1 (N/A) charge2 (N/A) charge1 – Ratio between sigmaCharge (( ሻ π‘‘β„Žπ‘π‘ π‘•π‘“ βˆ’ π‘›π‘“π‘π‘œπ·β„Žπ‘π‘ π‘•π‘“ 2ሻ and the mean charge in collection plane. charge2 – Ratio of charge in the last 10% of the PFO and the mean charge in the collection plane

11

slide-12
SLIDE 12

Current Approach In Pandora

Cut flow approach to track/shower characterisation

12

slide-13
SLIDE 13

Cut-flow approach

  • Does some basic cuts based on topological variables and then decide

if a PFO is track-like or shower-like

  • For anyone interested (PfoCharacterisationBaseAlgorithm.cc)

13

slide-14
SLIDE 14

BDT1, BDT2, and BDT3

14

slide-15
SLIDE 15

BDT setup

BDT1

  • Only uBooNE variables
  • Sensible fiducial cuts (20cm in x and y

direction, and 200 cm in z direction)

  • Completeness >= 0.1 && Purity >= 0.5
  • 50% numu and 50% nue MCC11

samples

  • Veto mischaracterised pfo (will talk

about it in the next slide)

  • After cuts -> 137,000 signals (track-

like) and 52,000 backgrounds (shower-like)

  • kBDT, Ntrees = 800, MinNodeSize=5%,

MaxDepth=3, BoostType=AdaBoost, AdaBoostBeta=0.5, BaggedSampleFraction=0.6, SeparationType=GiniIndex, nCuts=20 BDT2

  • Same as BDT1
  • Completeness >= 0.8 && Purity >= 0.8
  • After cuts -> 52,000 signals (track-like)

and 23,000 backgrounds (shower-like) BDT3

  • Same as BDT2
  • uBooNE variables + hierarchy

variables (will talk about it later as well)

15

slide-16
SLIDE 16

mischaracterisedPfo

Motivation Description

  • For a given PFO, work out what

MCParticle each hit maps to

  • Workout showerProbability where,
  • π‘‘β„Žπ‘π‘₯π‘“π‘ π‘„π‘ π‘π‘π‘π‘π‘—π‘šπ‘—π‘’π‘§ = π‘’π‘π‘’π‘π‘š π‘‘β„Žπ‘π‘₯π‘“π‘ βˆ’π‘šπ‘—π‘™π‘“ β„Žπ‘—π‘’π‘‘ π‘—π‘œ π‘π‘šπ‘š 𝑀𝑗𝑓π‘₯

π‘’π‘π‘’π‘π‘š β„Žπ‘—π‘’π‘‘ π‘—π‘œ π‘π‘šπ‘š 𝑀𝑗𝑓π‘₯

  • Flag a PFO as mischaracterised if

showerProbability >= 0.5 but called as track-like or vice versa

  • Use this flag to improve BDT training
  • Essentially, don’t train on wrong

topology

NCDIS_P_P_P_PIZERO

16

slide-17
SLIDE 17

Examples of vetoed and remaining PFOs

Mischaracterised Tracks Mischaracterised Showers Remaining Showers Remaining Tracks

17

slide-18
SLIDE 18

BDT value distribution for BDT1

BDT value Tracks

18

slide-19
SLIDE 19

BDT value distribution for BDT2

BDT value Tracks

19

slide-20
SLIDE 20

Hierarchy Variables

What are they?

20

slide-21
SLIDE 21

Hierarchy variables

  • Indicates hadronic activities
  • 3 new variables
  • nAllDaughter
  • nHits3DDaughterTotal
  • daughterParentNhitsRatio
  • Aim is to push protons and pions towards the track-like region to

improve track/shower separation

21

slide-22
SLIDE 22

nHits3DDaughterTotal nAllDaughter – total number of all downstream daughter pfos nHits3DDaughterTotal – total number of 3D hits in all downstream daughter pfos nAllDaughter

22

slide-23
SLIDE 23

daughterParentNHitsRatio daughterParentNHitsRatio daughterParentNhitsRatio – 3D hits ratio between all downstream daughter pfos and parent pfo.

23

slide-24
SLIDE 24

BDT value

BDT value distribution for BDT3

Tracks

24

slide-25
SLIDE 25

Comparing DifferentBDTs

100 95 90 85 80 75 70 PIPLUS

  • CCDIS_MU_PIPLUS
  • (2047)

PIMINUS - NCDIS_P_PIMINUS - (988)

CorrectID

E L E C T R O N

  • CCQ

EL_E_P - (3319) PHOTON - NCRES_PIZERO

  • (1184) PROTON - CCRES_E_P_PIPLUS-

MUON - CCQEL_MU_P - (3386) (1956)

Particle - Interaction -(#events)

Current Pandora BDT2 BDT1 BDT3 25

slide-26
SLIDE 26

Validation Output - Electron

Current Pandora BDT3 CCQEL_E

  • ELECTRON: nEvents: 3419, correctId 92.478%

CCQEL_E

  • ELECTRON: nEvents: 3419, correctId 92.205%

CCQEL_E_P

  • ELECTRON: nEvents: 3319, correctId 91.1111%

CCQEL_E_P

  • ELECTRON: nEvents: 3319, correctId 92.9938%

CCRES_E_P_PIPLUS

  • ELECTRON: nEvents: 1956, correctId 89.6534%

CCRES_E_P_PIPLUS

  • ELECTRON: nEvents: 1956, correctId 91.0189%

CCRES_E_P_PIZERO

  • ELECTRON: nEvents: 822, correctId 92.2094%

CCRES_E_P_PIZERO

  • ELECTRON: nEvents: 822, correctId 92.848%

26

slide-27
SLIDE 27

Validation Output - Photon

Current Pandora BDT3 CCRES_E_P_PIZERO

  • PHOTON1: nEvents: 822, correctId 77.4576Z

CCRES_E_P_PIZERO

  • PHOTON1: nEvents: 822, correctId 80.5085%

CCDIS_MU_P_PIZERO

  • PHOTON1: nEvents: 981, correctId 87.6457%

CCDIS_MU_P_PIZERO

  • PHOTON1: nEvents: 981, correctId 90.9091%

NCRES_PIZERO

  • PHOTON1: nEvents: 1184, correctId 75.5474%

NCRES_PIZERO

  • PHOTON1: nEvents: 1184, correctId 71.2591%

NCRES_P_PIZERO

  • PHOTON1: nEvents: 834, correctId 79.3829%

NCRES_P_PIZERO

  • PHOTON1: nEvents: 834, correctId 81.9074%

27

slide-28
SLIDE 28

Validation Output - Muon

Current Pandora BDT3 CCQEL_MU

  • MUON: nEvents: 3161, correctId 99.6145%

CCQEL_MU

  • MUON: nEvents: 3161, correctId 99.8394%

CCRES_MU_P

  • MUON: nEvents: 3386, correctId 99.3714%

CCQEL_MU_P

  • MUON: nEvents: 3386, correctId 99.8503%

CCRES_MU_P_PIPLUS

  • MUON: nEvents: 1875, correctId 98.75%

CCRES_MU_P_PIPLUS

  • MUON: nEvents: 1875, correctId 99.1848%

CCDIS_MU_PIPLUS

  • MUON: nEvents: 2047, correctId 97.6744%

CCDIS_MU_PIPLUS

  • MUON: nEvents: 2047, correctId 98.8878%

28

slide-29
SLIDE 29

Validation Output - Proton

Current Pandora BDT3 CCQEL_MU_P

  • PROTON1: nEvents: 3386, correctId 94.0557%

CCQEL_MU_P

  • PROTON1: nEvents: 3386, correctId 98.7307%

CCQEL_E_P

  • PROTON1: nEvents: 3319, correctId 94.2895%

CCQEL_E_P

  • PROTON1: nEvents: 3319, correctId 98.4903%

CCRES_E_P_PIPLUS

  • PROTON1: nEvents: 1956, correctId 95.2663%

CCRES_E_P_PIPLUS

  • PROTON1: nEvents: 1956, correctId 98.4615%

CCRES_MU_P_PIPLUS

  • PROTON1: nEvents: 1875, correctId 95.3071%

CCRES_MU_P_PIPLUS

  • PROTON1: nEvents: 1875, correctId 98.5516%

29

slide-30
SLIDE 30

Validation Output - PiPlus

Current Pandora BDT3 CCRES_MU_PIPLUS

  • PIPLUS: nEvents: 1867, correctId 83.4203%

CCRES_MU_PIPLUS

  • PIPLUS: nEvents: 1867, correctId 92.6377%

CCRES_E_P_PIPLUS

  • PIPLUS: nEvents: 1956, correctId 78.7336%

CCRES_E_P_PIPLUS

  • PIPLUS: nEvents: 1956, correctId 91.3979%

CCDIS_MU_PIPLUS

  • PIPLUS: nEvents: 2047, correctId 86.2016%

CCDIS_MU_PIPLUS

  • PIPLUS: nEvents: 2047, correctId 91.6796%

CCDIS_MU_P_PIPLUS

  • PIPLUS: nEvents: 825, correctId 85.3247%

CCDIS_MU_P_PIPLUS

  • PIPLUS: nEvents: 825, correctId 90.6494%

30

slide-31
SLIDE 31

Validation Output - PiMinus

Current Pandora BDT3 NCDIS_P_PIMINUS

  • PIMINUS: nEvents: 988, correctId 89.8361%

NCDIS_P_PIMINUS

  • PIMINUS: nEvents: 988, correctId 90.7104%

NCRES_P_PIMINUS

  • PIMINUS: nEvents: 574, correctId 88.417%

NCRES_P_PIMINUS

  • PIMINUS: nEvents: 574, correctId 95.1737%

31

slide-32
SLIDE 32

What’s next?

  • Replicate these results inside Pandora using SKLearn
  • Do more tests
  • Release
  • Deep Learning approaches

32

slide-33
SLIDE 33

The End

Thank you. Any questions?

33

slide-34
SLIDE 34

Backup Slides

34

slide-35
SLIDE 35

Small Note On Completeness VS Purity

35

slide-36
SLIDE 36

The GIF below cycles from Completeness && Purity >= 0.0 to 1.0.

36

slide-37
SLIDE 37

The GIF below cycles from Purity >= 0.5 && Completeness >= 0.0 to 1.0.

37

slide-38
SLIDE 38

The GIF below cycles from Completeness >= 0.5 && Purity >= 0.0 to 1.0.

38

slide-39
SLIDE 39

Correlation Matrix for 13 variables in BDT3

39

slide-40
SLIDE 40

How removing nHits3DDaughterTotal affects Track/Shower ID

BDT3 w/o nHit3DDaughterTotal BDT3

40

slide-41
SLIDE 41

nHits3DDaughterTotal for Track-like particles

41