AI/ML for Particle Accelerator, X-Ray Beamlines and Electron Microscopy

1 Nov 2021, 10:00 → 3 Nov 2021, 16:00 US/Central

Virtual

AI for Particle Accelerators, X-ray Beamlines, and Electron Microscopy Workshop @ ANL

Advances in instrumentation have dramatically increased the complexities associated with  experimental facilities. This includes enhanced facility capabilities as well as a substantial increase in the data generated. Consequently, the control and diagnostics of these experimental facilities are becoming increasingly complex, and the large output data streams necessitate smarter and more automated management and analyses of the data. Artificial Intelligence (AI) methods hold the promise of substantially improved management, control, and data analyses with the potential to dramatically increase experimental efficiencies as well as expanding and accelerating scientific discoveries.

Argonne is the home to world-leading facilities such as the Advanced Photon Source (APS), the Argonne Tandem Linear Accelerator (ATLAS), the Argonne Wakefield Accelerator (AWA), and the Electron Microscopy Center at the Center for Nanoscale Materials (CNM). In order to highlight AI opportunities in these facilities, Argonne is hosting a workshop on AI for with participants drawn from 3 communities: particle accelerators, X-ray beamlines and electron microscopy. The goals of the workshop are:

• review problems in control,  diagnostics, and data management that can be remedied by AI tools
• explore commonalities in problems and potential AI solutions across the 3 communities: particle accelerators, X-ray beamlines, and electron microscopy
• stimulate interactions across the communities to better leverage expertise and resources.

The workshop will be held virtually Nov. 1 – 3, 2021. The topics covered will be:

• Tuning and Control
• Data Analytics
• Imaging and Data Management
• Failure Detection, Virtual Diagnostics, and Digital Twins
• Autonomous Discovery
• Data, Computing, and Modeling

Each topic will open with invited presentations that give an overview of the current state, and set the stage for discussions – the discussions are the central part of the workshop and we will encourage broad and lively participation in the discussions.

The workshop is open to all of ANL. Registration information will be announced shortly.

WORKSHOP ORGANIZERS FROM EACH OF THE THREE COMMUNITIES

Particle Accelerators	X-ray Beamlines	Electron Microscopy
Michael Borland (APS)  Brahim Mustapha (PHY)  John Power (HEP)	Olle Heinonen (PSE)  Nicholas Schwarz (APS)  Martin Holt (NST)	Jianguo Wen (NST)  Charudatta Phatak (MSD)

Monday, 1 November

10:00 → 12:15 Opening Session: Tuning and Control

Convener: Michael Borland (Argonne National Laboratory)

10:00 Opening Remarks

Speakers: Brahim Mustapha (Argonne National Laboratory), Olle Heinonen (Argonne National Laboratory)

AI-ML-Workshop-Opening-Remarks.pdf

10:05 Online tuning of particle accelerators with reinforcement learning and comprehensive surrogate models

Speaker: Auralee Edelen (SLAC)

20211101_AIML_ANL_Edelen.pdf

A.Edelen_AIML@ANL.pdf

10:30 Automated Experiments, Machine Learning and Reinforcement Learning for tuning and control applications in microscopy

Speaker: Rama Vasudevan (Oak Ridge National Laboratory)

rkv_argonne.pdf

R.Vasudevan_AIML@ANL.pdf

10:55 Bayesian Techniques for Accelerator Characterization and Control

Speaker: Ryan Roussel (University of California Los Angeles)

Roussel_AWA_AI_ML_workshop.pdf

R.Roussel_AIML@ANL.pdf

11:20 Simultaneous online optimization of dynamic acceptance and lifetime for APS

Speaker: Yipeng Sun (Argonne National Laboratory)

YipengSun-2021Nov.pdf

11:40 Open / Speaker Panel Discussion

12:15 → 13:30 Lunch Break

13:30 → 15:45 Session II: Data Analytics

Convener: Martin Holt (ANL)

13:30 Accelerating atomic-scale characterization by combining theory and AI/ML

Speaker: Maria Chan (Argonne National Laboratory)

M.Chan_AIML@ANL.pdf

13:55 Solving X-ray and Cryo-EM Inverse Problems with Deep Learning

Speaker: Daniel Ratner (SLAC)

Argonne_InverseProblems_Nov2021.pptx

D.Ratner_AIML@ANL.pdf

14:20 Supervised and Unsupervised Machine Learning for Large, Noisy STEM Data

Speaker: Paul Voyles (University of Wisconsin - Madison )

ML for STEM ANL AI Workshop 2021.pdf

P.Voyles_AIML@ANL.pdf

14:45 Automated x-ray PDF analysis and ML modeling

Speaker: Chris Benmore (Argonne National Laboratory)

Benmore_ML_X-ray_Oct2021.pdf

C.Benmore_AIML@ANL.pdf

15:10 Open / Speaker Panel Discussion

Tuesday, 2 November

10:00 → 12:30 Session III: Failure Detection, Virtual Diagnostics & Digital Twin

Convener: John Power (Argonne National Lab)

10:00 Applications of Neural Networks for Anomaly Detection in Particle Accelerators

Speaker: Jonathan Edelen (RadiaSoft LLC)

J.Edelen_AIML@ANL.pdf

J.Edelen_AIML@ANL_Talk.pdf

10:25 Data-Driven Detection, Isolation, Identification, and Prediction of Accelerator Fault Events

Speaker: Chris Tennant (Jefferson Lab)

ANL_TENNANT_11022021.pdf

C.Tennant_AIML@ANL.pdf

10:50 Adaptive Machine Learning for Control and Virtual Diagnostics of Time-Varying Particle Accelerator Systems and Beams

Speaker: Alexander Scheinker (Los Alamos National Laboratory)

A.Scheinker_AIML@ANL.pdf

Scheinker_ANL_2021_vFinal.pdf

11:15 A Digital Twin for Spatiotemporal Experiments

Speaker: Subramanian Sankaranarayanan (Argonne National Laboratory)

S.Sankara_AIML@ANL.pdf

11:40 Uncertainty aware anomaly detection to predict errant beam pulses in the SNS accelerator

Speaker: Malachi Schram (Thomas Jefferson National Accelerator Facility)

M.Schram_AIML@ANL.pdf

M.Schram_AIML@ANL_Talk_v4.pdf

11:55 Open / Speaker Panel Discussion

12:30 → 13:30 Lunch Break

13:30 → 15:45 Session IV: Imaging and Data Processing

Convener: Charudatta Phatak (ANL)

13:30 Using automated image analysis to understand nanoparticle coarsening, ripening and evaporation at the nanoscale

Speaker: Eric Stach (University of Pennsylvania)

13:55 Phase retrieval with deep learning network and automatic differentiation

Speaker: Tao Zhou (Argonne National Laboratory)

T.Zhou_AIML@ANL.pdf

14:20 Overview of AI/ML@APS beamlines

Speaker: Mathew Cherukara (Argonne National Laboratory)

14:45 Accelerated Data Analysis Workflows for Ptychography using Remote Leadership Computing Resources

Speaker: Tekin Bicer (Argonne National Laboratory)

T.Bicer_AIML@ANL.pdf

15:10 Open / Speaker Panel Discussion

Wednesday, 3 November

10:00 → 12:15 Session V: Autonomous Discovery

Convener: Jianguo Wen (ANL)

10:00 Imaging 2D materials, atom by atom, on the million-atom scale

Speaker: Pinshane Huang (University of Illinois Urbana-Champaign)

P.Huang_AIML@ANL.pdf

10:25 Recent Progress in Autonomous X-ray Scattering

Speaker: Kevin Yager (Brookhaven National Laboratory)

K.Yager_AIML@ANL.pdf

10:50 Domain-Aware Gaussian Processes and High-Performance Mathematical Optimization for Stochastic Function Approximation and Autonomous Data Acquisition

Speaker: Marcus Noack (Lawrence Berkeley National Laboratory)

M.Noack_AIML@ANL.pdf

11:15 Advanced Robotics and Software Defined Automation to Speed Materials Discovery

Speaker: Ben Blaiszik (University of Chicago)

11:40 Open / Speaker Panel Discussion

12:15 → 13:30 Lunch Break

13:30 → 15:45 Closing Session: Data, Computing and Modeling

Convener: Nicholas Schwarz (APS / ANL)

13:30 Programmatic and Deep Learning Analysis Pipelines for 4D-STEM Materials Science Experiments

Speaker: Colin Ophus (Lawrence Berkeley National Laboratory)

C.Ophus_AIML@ANL.pdf

13:55 MLExchange, bringing machine learning to the beamline

Speaker: Alexander Hexemer (Lawrence Berkeley National Laboratory)

A.Hexemer_AIML@ANL.pdf

14:20 The FNAL accelerator complex initial ML efforts and a conceptual AI roadmap for operations

Speaker: Tia Miceli (NMSU)

FNAL_for_Argonne_workshop.pdf

T.Miceli_AIML@ANL.pdf

14:45 Surrogate Models for Beam Transport and Particle Tracking

Speaker: Jose Martinez (Argonne National Laboratory)

J.Martinez_AIML@ANL.pdf

15:05 Open / Speaker Panel Discussion

15:40 Closing Remarks

Speakers: Brahim Mustapha (Argonne National Laboratory), Olle Heinonen (Argonne National Laboratory)