Mini-workshop on lattice gauge theory and related topics for high-energy physics

Wednesday 14 May 2025, 09:00 → 17:30 Asia/Taipei

Lecture room 6 (NYCU Beimen campus)

                              

This mini-workshop aims to bring together experts in various topics related to lattice gauge theory and their applications to problems in high-energy physics.

Invited speakers and their talk titles are

• Guilherme Catumba (University of Milano)
  Tackling the Signal to Noise Problem with Stochastic Automatic Differentiation
   
• Miranda C.N. Cheng (Academia Sinica and University of Amsterdam)
  TBA
   
• Luigi Del Debbio (University of Edinburgh)
  Machine Learning for Parton Distribution Functions 
   
• Mathis Gerdes (University of Amsterdam)
  TBA
   
• Wayne Morris (National Yang Ming Chiao Tung University)
  TMD Soft Function on the Lattice Using Complex Directional Wilson Lines
   
• Alberto Ramos (University of Valencia)
  TBA
   

All are welcome.

 

09:30 → 09:45 Opening

Speaker: Prof. C.-J. David Lin (National Yang Ming Chiao Tung University)

09:45 → 10:30 Machine Learning for Parton Distribution Functions

The NNPDF collaboration has been using Machine Learning techniques to solve the inverse problem of extracting Parton Distribution Functions from finite sets of experimental data for almost two decades. With the increased precision of the LHC measurements, It has become mandatory to understand the robustness of the error bars and of the correlations in the results of PDFs fit. We review the fitting procedure from a Bayesian perspective and discuss the training of neural networks in a Bayesian framework.

Speaker: Prof. Luigi Del Debbio (University of Edinburgh)

10:30 → 11:00 Coffee

11:00 → 11:45 TMD soft function on the lattice using complex directional Wilson lines

Speaker: Dr Wayne Morris (National Yang Ming Chiao Tung University)

11:45 → 12:30 Tackling the Signal to Noise problem with Stochastic Automatic Differentiation

Lattice field theory computations of two-point functions are generally
affected by the so called signal to noise problem, wherein the signal
of the Euclidean time correlator decays faster than the variance. In
this talk we propose a different perspective on the origin of this
problem. Following this, we argue that by writing correlators as
derivatives with respect to sources and evaluating these derivatives
using techniques of stochastic automatic differentiation we can
eliminate completely the signal to noise problem. Results in a four
dimensional scalar theory confirm the expected behavior.

Speaker: Dr Guilherme Catumba (University of Milano)

12:30 → 14:00 Lunch

14:00 → 14:45 TBA

14:45 → 15:30 TBA

15:30 → 16:00 Coffee

16:00 → 16:45 TBA

16:45 → 17:30 Discussion