Summer Institute on Wigner Imaging and Femtography Simonetta Liuti - - PowerPoint PPT Presentation

▶
summer institute on wigner imaging and femtography
SMART_READER_LITE
LIVE PREVIEW

Summer Institute on Wigner Imaging and Femtography Simonetta Liuti - - PowerPoint PPT Presentation

Mar 07, 2024 253 likes •635 views

8/7/2019 siwif-logo.png Summer Institute on Wigner Imaging and Femtography Simonetta Liuti University of Virginia CNF Symposium August 12-13, 2019 SURA Headquarters, Washington DC 2 8/13/19 Summer Institute for Wigner Imaging and

slide-1
SLIDE 1

Simonetta Liuti University of Virginia CNF Symposium August 12-13, 2019 SURA Headquarters, Washington DC

Summer Institute on Wigner Imaging and Femtography

8/7/2019 siwif-logo.png

slide-2
SLIDE 2

8/13/19

2

Summer Institute for Wigner Imaging and Femtography

Wigner Theory

Data Management/ Communication

Observables

Machine Learning

Simonetta Liuti

Principle Investigator University of Virginia

Librado Anglero

University of Virginia Physics

Fatma Aslan

New Mexico State University PhD

Emma Yeats

University of Virginia Physics

Kyle-Thomas Pressler

University of Virginia Physics

Fernanda Yepez-Lopez

University of Virginia Mathematics

Pete Alonzi

Co Principle Investigator University of Virginia

Yao(Grace) Tong

University of Virginia Mathematics and Economics

Brandon Kriesten

University of Virginia Observables Group Leader

Meg Graham

University of Virginia Computer Science

Jake Grigsby

Machine Learning Group Leader University of Virginia Computer Science and Mathematics

Evan Anders Magnusson

University of Virginia Computer Engineering and Computer Science

Christopher Thompson

Virginia Union University Physics and Engineering

Krisean D Allen

Virginia Union University Physics

Andrew Meyer

University of Virginia

William A Oliver

Virginia Commonwealth University

Yelena Prok

Virginia Commonwealth University Assistant Professor

Matthias Burkardt

Co Principle Investigator New Mexico State University

Consultant

Dustin Keller

Co Principle Investigator University of Virginia

Olivier Pfister

Co Principle Investigator University of Virginia

Carlos Gonzalez Arciniegas

University of Virginia

Timothy John Hobbs

Southern Methodist University EIC Center at Jefferson Lab

Gabriel Niculescu

James Madison University

Abha Rajan

University of Virginia

*** Red: Undergraduate Blue: Graduate

slide-3
SLIDE 3

8/13/19

3

Our project was characterized from the very beginning by a great response from students and young researchers:

slide-4
SLIDE 4

8/13/19

4

3 contributions submitted to CEU@ Annual Fall Meeting of DNP!

https://www.uwlax.edu/ceu/current/

slide-5
SLIDE 5

8/13/19

5

A truly interdisciplinary effort

3D Structure of the proton

Data Science

Quantum Information Education Outreach

slide-6
SLIDE 6

Main Questions we want to answer

8/13/19

6

What are the necessary steps for an unbiased extraction of the 3D structure and mechanical properties of the proton? GPDs and Compton Form Factors are the observables Probing nuclear physics with gravitational waves

slide-7
SLIDE 7

8/13/19

7

Wigner Theory

Modeling the Wigner Function Exploring issues in common with Atomic Physics

Data Analysis

Modeling the Cross Section Extracting Observables from Data Error Analysis

Visualization

Outreach/ pedagogical Research Tool

Communication

Data Management Dissemination Webpage/github Code sharing

  • pensourcing
slide-8
SLIDE 8

8/13/19

8

Wigner Theory Modeling the Wigner Function Exploring issues in common with Atomic Physics

slide-9
SLIDE 9

8/13/19

9

ep → e'γ ' p'

e’ k’ k e p’ p

  • (k-k’)2=Q2

q’

Deeply Virtual Compton Scattering

large scale

slide-10
SLIDE 10

8/13/19

10

We focus on GPDs (one projection of the Wigner distribution which is directly observable in DVCS) The observables are the Compton Form Factors which are convolutions of GPDs with known kernels

slide-11
SLIDE 11

Evaluate Observables d5σDVCS ALU, ALL, AUT,… Constraint 2 PDFs u(x,Q2), u(x,Q2), … Δu(x,Q2), Δd(x,Q2), G(x, Q2) Data DVCS,TCS,DVMP Constraint 1 Nucleon Form Factors F1, F2,GA, GV Evaluate CFFs 2nd partial parameters fix Evaluate Estimator Solve pQCD Evolution Fit Result 1st partial parameter fix Generate Wigner/GPD forms @ Qo

2

Hi=u,d,..({a}), Ei=u,d,..({a}),…

Flowchart/roadmap from data/observables to GPDs

slide-12
SLIDE 12

8/13/19

12

Hi=u,d,...(x, ξ, t) = NG{a}(x, ξ, t) R{β}(x, ξ, t) {a} = MX, m, MΛ {β} = α, α0, p

from PQCD evolved GPD H(x,0,t) (parametrization from J. O. Gonzalez et al., arXiv:1206.1876, PRC88(2013) ) …to Fourier transform: H(x,b) parameter variations used for input for Data Analysis

H(x,0,t) x Fixed t

u,d quarks and gluon GPDs

slide-13
SLIDE 13

8/13/19

13

Working on Connection with Atomic Physics/Quantum Information

with O. Pfister and C. Gonzalez Arciniegas

slide-14
SLIDE 14

8/13/19

14

Experimental Results from UVA

  • C. Gonzalez Arciniegas

https://pages.shanti.virginia.edu/Femtography/ files/2019/06/Seminar.pdf

slide-15
SLIDE 15

8/13/19

15

Data Analysis Modeling the Cross Section Extracting Observables from Data Error Analysis

slide-16
SLIDE 16

8/13/19

16

Analytic Development: Precise Formulation of Cross Section Machine Learning correlated work I n t e r d i s c i p l i n a r y e f f

  • r

t w i t h d a t a s c i e n c e

slide-17
SLIDE 17

8/13/19

17

  • What physics questions drive our choice?
  • What type of Machine Learning?
  • What kinematical domains are sensitive to which

GPDs/Compton Form Factors

  • With what uncertainty?
slide-18
SLIDE 18

Evaluate Observables d5σDVCS ALU, ALL, AUT,… Constraint 2 PDFs u(x,Q2), u(x,Q2), … Δu(x,Q2), Δd(x,Q2), G(x, Q2) Data DVCS,TCS,DVMP Constraint 1 Nucleon Form Factors F1, F2,GA, GV Evaluate CFFs 2nd partial parameters fix Evaluate Estimator Solve pQCD Evolution Fit Result 1st partial parameter fix Generate Wigner/GPD forms @ Qo

2

Hi=u,d,..({a}), Ei=u,d,..({a}),…

Flowchart/roadmap from data/observables to GPDs ML TensorFlow (logo

slide-19
SLIDE 19

8/13/19

19

Modeling the Cross Section Introducing the complete formalism where all kinematical dependences are precisely known and approximations are under control

slide-20
SLIDE 20

8/13/19

20

d5σBH

unpol

dxBjdQ2d|t|dφdφS = Γ t2 ⇥ ABH

  • F 2

1 + τF 2 2

  • + BBHτG2

M(t)

⇤ d5σI

unpol

dxBjdQ2d|t|dφdφS = Γ Q2(t) h AI (F1<eH + τF2<eE) + BI GM <e(H + E) + CI

Rosenbluth separation for Bethe-Heitler contribution

slide-21
SLIDE 21

8/13/19

21

d5σBH

unpol

dxBjdQ2d|t|dφdφS = Γ t2 ⇥ ABH

  • F 2

1 + τF 2 2

  • + BBHτG2

M(t)

⇤ d5σI

unpol

dxBjdQ2d|t|dφdφS = Γ Q2(t) h AI (F1<eH + τF2<eE) + BI GM <e(H + E) + CI GM <e e H i

G GM

2

GM GA GE

2

slide-22
SLIDE 22

8/13/19

22

Extracting Observables from Data/Error Analysis The goal is to single out what measurements have the greatest impact on determining the Compton Form Factors. To address this challenge we present two analysis methods to identify both the type of experiments (beam/target polarization) and the kinematical domain providing the best constraints.

slide-23
SLIDE 23

8/13/19

23

Based on Supervised Learning…

slide-24
SLIDE 24

8/13/19

24

Strategy:

  • 1. A fully connected neural

network maps input kinematic data to a vector

  • f eight form factors (see

diagram).

Femtography Imaging with Neural Networks (FINN)

We translate the x-sec. code into TensorFlow Automatically differentiable At variance with other efforts we can train CFF extraction network with backpropagation and variants

  • f stochastic gradient descent.

1.

  • 2. Use a code developed by
  • ur Data Analysis Team to

evaluate the cross sections and in terms of the CFFs.

slide-25
SLIDE 25

8/13/19

25

present data kinematical coverage Understanding the Error/Systematic bias: example xBj Q2

Small error Large error

Re H

slide-26
SLIDE 26

8/13/19

26

The biggest challenge we have is limited data, which we can solve with a combination of regularization, Monte Carlo generation and interpolation between phi values for each kinematic range. Larger datasets from future DVCS/TCS experiments will ease these engineering challenges.

slide-27
SLIDE 27

8/13/19

27

Based on unsupervised learning….

slide-28
SLIDE 28

8/13/19

28

Large t diquark correlations Small t Reggeized diquark model Ultimately, we want to understand features of the GPDs behavior

slide-29
SLIDE 29

8/13/19

29

Using UMAP (Uniform Manifold Approximation and Projection) to cluster GPDs from the Wigner Theory team into 2D Can be extended to 3D for virtual reality!

slide-30
SLIDE 30

https://arxiv.org/pdf/1803.02777.pdf

Principal Component Analysis

… similarities with CTEQ analysis

slide-31
SLIDE 31

8/13/19

31

Visualization Outreach/ pedagogical Research Tool

slide-32
SLIDE 32

8/13/19

32

Outreach tool

slide-33
SLIDE 33

8/13/19

33

Stab at visualizing UMAP

slide-34
SLIDE 34

8/13/19

34

Communication Data Management Plan Dissemination Webpage/github Code sharing Open-sourcing See Pete Alonzi’s talk on Monday

slide-35
SLIDE 35

8/13/19

35

Summary

We developed a strategy and specifically ML tools for analyzing deeply virtual exclusive experiments within a truly interdisciplinary effort Visualization tool for outreach Impact on Education: many undergradauate students from various departments involved in research, CEU presentations Several manuscripts in progress Website and model for both fostering and regulating community interactions

slide-36
SLIDE 36

8/13/19

36

Future Work Issues with data: situations where data are scarce, how to add in dynamically new sets of data, Including lattice QCD results/constraints Fourier transforms More work to come on Wigner Distributions and Quantum Information Developing Unsupervised Learning tool Visualization tools for research and more animations/movies for outreach tool

slide-37
SLIDE 37

8/13/19

37