1. WRM for Neutrinos - INTRODUCTION Dorota Stefan and Robert Sulej

2. Dorota Stefan, Robert Sulej
DUNE – Deep Underground Neutrino Experiment
neutrino beam
near detector
4 LArTPC detectors
SURF G4 G4
TDR
TDR
15m
15m
6 m
2 m
62m
Simulated longitudinal and transverse containment for proton showers of 4 and 10 GeV/c momenta.
Dorota Stefan, Robert Sulej

3. Dorota Stefan, Robert Sulej
DUNE – physics goals
Neutrinos:
1. CPV in leptonic sector.
2. n mass ordering.
3. Sterile neutrino.
4. Atmospheric neutrinos.
5. Neutrino astrophysics, including supernova n bursts.
6. Neutrino cross sections.
Barion number violation.
will interact somewhere in the detector (random location of primary interaction)
ne CC
n iteraction vertex
electron
LArSoft, Genie, G4, simulated neutrino interaction in DUNE far detector
Dorota Stefan, Robert Sulej

4. Data handling becomes a challenge
LArTPC events
low energy deposition events
DUNE, supernova ev.
High granularity detector:
3/5 wire mm pitch
Dense sampling: 0.6 mm
2/3/4 wire planes
20+ year operation with huge detector
LArSoft
LArSoft
high energy deposition events
DUNE, neCC
drift time
Data handling becomes a challenge
Wire number
Genie, G4, LArSoft
Dorota Stefan, Robert Sulej

5. Dorota Stefan, Robert Sulej
LArTPC events & region of interest
low energy deposition events
DUNE, supernova ev.
X [cm]
Y[cm]
Selected regions used for the offline processing
LArSoft
LArSoft
Z [cm]
high energy deposition events
DUNE, neCC
drift time
X [cm]
Y[cm]
Wire number
Genie, G4, LArSoft
Genie, G4, LArSoft
Z [cm]
Dorota Stefan, Robert Sulej

6. Dorota Stefan, Robert Sulej
Challenges in finding region of interest
Simple zero suppression based on 1D waveform analysis is very inefficient in LArTPC.
Obviously need to look at 2D correletions
Challenging due to:
variety of (changing) noise patterns,
number of different signal classes.
ICARUS, D. Stefan NuFACT11 Conf
Shower
Cosmic muon
Neutrino2016,
Jonathan Davies for DUNE Collaboration
Karl Warburton
IOP HEPP&APP Conf.
35-ton
35-ton
Dorota Stefan, Robert Sulej

7. Ali Abdallah, Roberto Cardarelli, Giulio Aielli, Marco Manca
WRM For Neutrinos
Ali Abdallah, Roberto Cardarelli, Giulio Aielli, Marco Manca

8. Sensors produces BIG DATA
The problem
Sensors produces BIG DATA
A sensor is whatever is able to produce an observable encoding events in a given format
Sensors produce data faster than we can process and store
We need to interact with data to take decisions in real time
We need to operate as close a possible to the sensor
Physical World Sensor Data

9. Weighted Resistor Matrix - WRM
The WRM is an analog computer performing a regression on input data and providing an estimation of the best fit parameters and the associated likelihood
An interconnected repetitive resistive network subjected to an electrostatic potential provides the necessary elements for a regression-like computation:
The diffusion potential as a notion of distance
The superposition principle allows to add voltages to built a correlation estimator

10. Weighted Resistor Matrix - WRM
From: R. Cardarelli et al.; “On a very fast topological trigger”; NIM A324(1993)
A fast tracking trigger for LHC experiments
The ratio RW/RS determines the voltage diffusion rate
It is coupled to a parallel network
Performing the sum to test patterns

11. WRM for EDUSAFE
EDUSAFE [1] is a Marie Curie ITN project focusing on research into the use of Virtual and Augmented Reality (VR/AR) during planned and emergency maintenance in extreme environments.
[1] EDUSAFE Website
Expanding the WRM: from particle physics to computer vision -> It was born to detect linear correlations like particle tracks.
Original WRM (1995) was able to process 1Mpixel in ~1ms. With new technologies, the speed increases accordingly
The idea is to use the WRM with a tracking algorithm
The WRM device can play a hardware acceleration role
Feeds tracking algorithm with pre-extracted features. Allowing them to save valuable time.

12. WRM for Neutrinos: Particle Data Analysed as Images
Collected data is similar to a normal grayscale image
The idea is to use the WRM to filter out the noise
Crop then region of interest
New WRM layout is object of the present proposal, optimized for the neutrino trigger problem
We want to obtain a fast trigger and optimal zero suppression system!