Decay vertex ID using CNN for p K+ Aaron Higuera University of - - PowerPoint PPT Presentation

Decay vertex ID using CNN for p → K+

Aaron Higuera University of Houston

¯ ν

2

CNN Tools on the Proton Decay Analysis

Standard vertex reconstruction use pmtracks in order to reconstruct a vertex Ideally we would expect three two decay vertices K →µ Proton decay occurs at the nucleus

• nly FS particles are detectable in

the TPC (as in G4 simulation)

p → K+⊽ K+ µ+ e+ Standard vertex reco

3

Proton decay simulation

Time Time

p → K+⊽ → µ++vµ → e++ve K µ+ e+

CNN Tools on the Proton Decay Analysis

4

target class[had_int decay γConv] Classify the central pixel target class [decay] Previous model was trained using

CNN Tools on the Proton Decay Analysis

5

target class background (random point)

CNN Tools on the Proton Decay Analysis

6

CNN

target input (decay vertex) background (random point)

Model GPUs

convolutional layers dense layers

7

CNN Tools on the Proton Decay Analysis

Dump all hits of each track, then use the CNN model to predict hit w/vertex, select candidates above a pointID threshold then look for coincidence with space points and create a vertex hits w/ predicted vertex

p → K+⊽ K+ µ+ e+

8

CNN Tools on the Proton Decay Analysis

distance to true decay vertex (cm) 0.5 1 1.5 2 2.5 3 3.5 4 Arbitrary Units 0.05 0.1 0.15 0.2 distance to true decay vertex (cm) 0.5 1 1.5 2 2.5 3 3.5 4 Arbitray Units 0.05 0.1 0.15 0.2

is this better than standard reco?

Using CNN decay vertex Using end of the track Using CNN decay vertex Using end of the track

For Kaons For Muons

9

CNN Tools on the Proton Decay Analysis

How many vertices do we find? Efficiency for events with K → µ+⊽ Events with kaon & muon track (based PIDA) New model target class [decay] Old model target class [had_int, decay, γconv

10

CNN Tools on the Proton Decay Analysis

The challenges

K+ K =106 MeV (KE) K+ K+ heavy ionization hit accidentally selected as decay vtx Zoom In K+ µ+

11

CNN Tools on the Proton Decay Analysis

K+ µ+ e+ K =41 MeV (KE) p =23 MeV (KE) The model is blind to any decay that comes from kaons with KE below 50 MeV This is to avoid decay coming from nothing (particles with one or two hits)

There is a selection threshold

12

CNN Tools on the Proton Decay Analysis

K+ K+ K+ µ+ µ+ µ+

2 decay vertices examples

13

Comments

✤ Results keep improving ✤ More improvements are needed ✤ Vertex and track associations ✤ Suggestion?

14

The End

15

Workflow

1) Generate your sample background and signal 2) Dump the ADC info for further processing and training Output- This would produce adc map synced with pdg map 3) Prepare data input: patch ~20x20cm of deconvoluted ADC

• utput: vector of 3 values: [had_int decay pi0decay(gammaConv)]

4) Training your model GPU machine, thanks to Center for Advance Computing & Data Systems UH 5) Dump Keras model into a pure C++ model 6) Run ParticleDecayId_module.cc Uses hits and spacepoints from track with CNN info to produces vertices