AI4EIC 2023 Annual Workshop

US/Eastern
Caldwell auditorium (Catholic University of America, Washington D.C.)

Caldwell auditorium

Catholic University of America, Washington D.C.

Description

The 2023 annual workshop on AI4EIC --- Artificial Intelligence for the Electron Ion Collider, will be hosted by Catholic University of America in Washington, D.C., and will take place from November 28 to December 1, 2023.

This year's workshop will be a hybrid event with both in-person and virtual participation options. In light of the enriching experiences from the past two annual workshops at W&M and CFNS, we strongly encourage in-person attendance. Please note, the registration fee of $25 will be collected in person at the event and it is only for in-person participants. No fee is required for virtual participants.  (Registration at: https://indico.bnl.gov/event/19560/registrations/

The workshop will take place in the Caldwell auditorium in CUA.

Please find the Live Notes here

More details can be found at eic.ai/workshops. 

Proceedings will be published in the Journal of Instrumentation. The abstract submission (1 page max, figures can be included) is still open. 

Building upon the productive discussions and synergies formed during our previous events, the focus of this workshop will be the active and potential areas of AI/ML applications within the EIC, including ongoing activities in the ePIC experiment and beyond.

The workshop will feature AI/ML tutorial sessions led by experts from academia, national labs, and the industry.

We are also excited to announce the second international AI4EIC Hackathon, which will be held on December 1st.

AI/ML is proving to be a fundamental component in all phases of the future EIC, from design and R&D to practical applications. This workshop presents an invaluable opportunity for our community to stay abreast of the ongoing AI/ML projects and plans for the EIC. It's a platform for meaningful discussions on various interdisciplinary topics, fostering collaboration between theorists, experimentalists, accelerator physicists, and AI/ML practitioners.

 

Scientific Organizing Committee

Markus Diefenthaler (JLab)

Cristiano Fanelli (W&M)

Tanja Horn (CUA)

David Lawrence (JLab)

Meifeng Lin (BNL)

Dennis Manos (W&M)

Benjamin Nachman (LBNL)

Torre Wenaus (BNL)

Mike Williams (MIT)

 

Registration
3rd AI4EIC Workshop Registration Form
Participants
  • Abishek Karki
  • Adeel Mansha
  • Alamgir Khan
  • Alessandro Lonardo
  • Alexander Smith
  • Alexey Aparin
  • Andrea Cossu
  • Angela Marotta
  • Anil Panta
  • Anna Kowalewska
  • Anselm Vossen
  • Avnish Singh
  • Ayman Al-bataineh
  • Benjamin Nachman
  • Bhavya Singhal
  • Bianca Gualtieri
  • Biljana Mitreska
  • Brandon Kriesten
  • Brandon Tumeo
  • Cameron Dean
  • Carlos Yero
  • Casey Morean
  • Chandradoy Chatterjee
  • Charles Hughes
  • Churamani Paudel
  • Connor Pecar
  • Cristiano Fanelli
  • Daniel Murnane
  • David Lawrence
  • David Ruth
  • Derek Anderson
  • Dmitrii Kalinkin
  • Dmitry Romanov
  • Douglas Adams
  • Dulitha Jayakodige
  • Elena Fol
  • Elliott Kauffman
  • Eugenio scapparone
  • Evaristo Cisbani
  • Fidele Twagirayezu
  • Francis Atta Howard
  • Frederik Van der Veken
  • Gabriel Garcia
  • gabriele stagno
  • Gagik Gavalian
  • German Sborlini
  • Greg Kalicy
  • Gregory Matousek
  • Haile Eyasu Hagos
  • Haipeng Chen
  • Hemalata Nayak
  • Hrachya Marukyan
  • Huey-Wen Lin
  • Imtiaz Khan
  • Ishara Fernando
  • James Giroux
  • Jay Ajitbhai Sandesara
  • Jennefer Maldonado
  • Jesutofunmi Fajemisin
  • Ji Qiang
  • John Lajoie
  • Jonathan Edelen
  • Joshua Crafts
  • Jun Ma
  • karthik suresh
  • Kazuhiro Terao
  • Keith Jankowski
  • Kevin Brown
  • Kolja Kauder
  • Kyungseon JOO
  • Liliana Teodorescu
  • Liliet Calero Diaz
  • Luca Giommi
  • Lucy Lin
  • Lynn Wood
  • Maja Karwowska
  • Malachi Schram
  • MANAL ALMAEEN
  • Manouchehr Farkhondeh
  • Manuel Szewc
  • Maria Zurek
  • Matthew Kilpatrick
  • Matthew McEneaney
  • Maxim Potekhin
  • Maximiliano Sioli
  • MD IMRAN HOSSAIN
  • Meifeng Lin
  • mekaoui yahya
  • Mihai Mocanu
  • Mike Williams
  • Muhammad Farooq
  • Muhammad Junaid
  • Muhammad Junaid
  • Narbe Kalantarians
  • Nick Zachariou
  • Nicola Rubini
  • Nuwan Chaminda Gunawardhana Waduge
  • Patrick Moran
  • Raghav Kunnawalkam Elayavalli
  • Rahmat Rahmat
  • Rajeev Singh
  • Rick Stevens
  • Robert Edwards
  • Roberto Preghenella
  • Rowan Kelleher
  • Rui Zhang
  • Ryan Roussel
  • Saraswati Pandey
  • sayak chatterjee
  • shanjin wu
  • Shivam Raj
  • Simon Gardner
  • Simon Schneider
  • Simonetta Liuti
  • Stefano Iovieno
  • Steve Takouan Tchounga
  • Sudip Bhattarai
  • Sunil Pokharel
  • Talha Nazir
  • Tanja Horn
  • Tapasi Ghosh
  • Tareq Alghamdi
  • Thomas Britton
  • Timothy Hobbs
  • Tolga Erbora
  • Torre Wenaus
  • Torri Jeske
  • travon willis
  • travon willis
  • Tucker Hwang
  • Vinicius Mikuni
  • Wen Guan
  • Wenze Xi
  • William Phelps
  • Xiangyang Ju
  • Xiaofeng Gu
  • Yang Li
  • Yang-Ting Chien
  • Yaohang Li
  • Yeonju Go
  • Yi Huang
  • Yi-Kai Kan
  • Yong Du
  • Yue Hao
  • Zachary Baldwin
  • Zhenyu Ye
Contact
    • Welcome and Introduction
    • AI/ML for ePIC and Beyond
      Conveners: Anselm Vossen (member@duke.edu;faculty@duke.edu), Derek Anderson (Iowa State University)
      • 1
        Introduction: ePIC Overview
        Speaker: John Lajoie (Iowa State University)
      • 2
        Performance optimization for a scintillating glass electromagnetic calorimeter at EIC

        Successful realization of EIC scientific program requires designing
        and constructing high-performance particle detectors. Recent
        developments in the field of scientific computing and increased
        availability of high performance computing resources have made it
        possible to perform optimization of multi-parameter designs, even when
        the latter require longer computational times (for example simulations
        of particle interactions with matter). Procedures involving
        machine-assisted techniques used to inform the design decision have
        seen a considerable growth in popularity among the EIC detector
        community. Having already been realized for tracking and RICH PID
        detectors, it has a potential application in calorimetry designs. A
        SciGlass barrel calorimeter originally designed for EIC Detector-1 has
        a semi-projective geometry that allows for non-trivial perfomance
        gains, but also poses special challenges in the way of effective
        exploration of the design space while satisfying the available space
        and the cell dimension constraints together with the full detector
        acceptance requirement. This talk will cover specific approaches
        taken to perform this detector design optimization.

        Speaker: Dmitrii Kalinkin (University of Kentucky)
      • 3
        Object Condensation for Track Building in a Backward Electron Tagger at the EIC

        Quasi-real photoproduction measurements at the Electron Ion Collider will require a far backward electron tagger to detect electrons scattered at small angles close to the beam line. A high occupancy is expected in the electron tagger, leading to many possible permutations of hits deposited by the electrons into single electron tracks. To avoid a slow and computationally expensive combinatorial approach to track building, machine learning algorithms such as object condensation methods can be used to recognise objects such as tracks from hits in a detector. We demonstrate how these object condensation methods are particularly well suited to track building in the far backward electron tagger, achieving an efficiency in track finding at or above 95% and a purity at or above 90%, in the presence of noise and hit detection inefficiencies.

        Speaker: Dr Simon Gardner (University of Glasgow)
      • 11:05
        Coffee break
      • 4
        The Optimal use of Segmentation for Sampling Calorimeters

        One of the key design choices of any sampling calorimeter is how fine to make the longitudinal and
        transverse segmentation. To inform this choice, we study the impact of calorimeter segmentation
        on energy reconstruction. To ensure that the trends are due entirely to hardware and not to a sub-
        optimal use of segmentation, we deploy deep neural networks to perform the reconstruction. These
        networks make use of all available information by representing the calorimeter as a point cloud.
        To demonstrate our approach, we simulate a detector similar to the forward calorimeter system
        intended for use in the ePIC detector, which will operate at the upcoming Electron Ion Collider.
        We find that for the energy estimation of isolated charged pion showers, relatively fine longitudinal
        segmentation is key to achieving an energy resolution that is better than 10% across the full phase
        space. These results provide a valuable benchmark for ongoing EIC detector optimizations and may
        also inform future studies involving high-granularity calorimeters in other experiments at various
        facilities

        Speaker: Fernando Torales Acosta (Staff@lbl.gov;Member@lbl.gov;Other@lbl.gov)
      • 5
        AID(2)E: AI-Assisted Detector Design at EIC
        Speaker: Cristiano Fanelli (W&M)
      • 6
        Flash Talks
      • 7
        Discussion
    • 13:00
      Lunch
    • Calibration, Monitoring, and Experimental Control in Streaming Environments
      Conveners: Torri Jeske (Jefferson Lab), Yeonju Go (Brookhaven National Laboratory)
      • 8
        LHCb Calibration/Alignment

        In Run II of the LHC, a new scheme for the software trigger at LHCb allows splitting the triggering of events in two stages, allowing to perform the detector alignment and calibration in real time. The real-time alignment and calibration procedure is a fully automatic procedure at LHCb that is executed at the begining of each fill of the LHC. The alignment estimates the position of detector elements and the correct alignment contributes to achieve the best quality data for offline analysis. The procedure is implemented for the full tracking system at LHCb with the event reconstruction run as a multithreaded process. The procedure performs an update of the alignment constants, while the calibration constants are evaluated for each run. This allows identical constants to be used in the online and offline reconstruction and improves the correlation between triggered and offline selected events. The operational and technical point of view of this procedure during data-taking is discussed with the focus on performance and optimisations done regarding the new computing framework and the new detectors.

        Speaker: Biljana Mitreska (Dortmund Tech. U.)
      • 9
        ML-based Calibration and Control of the GlueX Central Drift Chamber

        The GlueX Central Drift Chamber (CDC) in Hall D at Jefferson Lab, used for detecting and tracking charged particles, is calibrated and controlled \textit{during} data taking using a Gaussian process. The system dynamically adjusts the high voltage applied to the anode wires inside the chamber in response to changing environmental and experimental conditions such that the gain is stabilized. Control policies have been established to manage the CDC's behavior. These policies are activated when the model's uncertainty exceeds a configurable threshold or during human-initiated tests during normal production running. We demonstrate the system reduces the time detector experts dedicate to calibration of the data offline, leading to a marked decrease in computing resource usage without compromising detector performance.

        Speaker: David Lawrence (Jefferson Lab)
      • 10
        Towards Fast Calibration with the ePIC Barrel Hadronic Calorimeter

        Measurement of jets and their substructure will provide valuable information about the properties of the struck quarks and their radiative properties in Deep-Inelastic Scattering events. The ePIC Barrel Hadronic Calorimeter (BHCal) will be a critical tool for such measurements. By enabling the measurement of the neutral hadronic component of jets, the BHCal will complement the Barrel Electromagnetic Calorimeter (BECal) and the ePIC tracking system to improve our knowledge of the jet energy scale. However, to obtain a physically meaningful measurement, the response of the combined BECal + BHCal system must be properly calibrated using information from both. With Machine Learning, this can be done in such a way that is both computationally efficient and easy to deploy in a production environment, making such an approach ideal for quasi-real time calibrations needed in a streaming readout environment. We present a potential machine-learning based algorithm for fast calibration of the combined system which could be deployed in such an environment and discuss progress towards its implementation.

        Speaker: Derek Anderson (Iowa State University)
      • 15:09
        Coffee break
      • 11
        Hydra: Computer Vision for Data Quality Monitoring

        Hydra is a system which utilizes computer vision to perform near real time data quality management, initially developed for Hall-D in 2019. Since then, it has been deployed across all experimental halls at Jefferson Lab, with the CLAS12 collaboration in Hall-B being the first outside of GlueX to fully utilize Hydra. The system comprises back end processes that manage the models, their inferences, and the data flow. The front-end components, accessible via web pages, allow detector experts and shift crews to view and interact with the system. This talk will give an overview of the Hydra system as well as highlight significant developments in Hydra's feature set, acute challenges with operating Hydra in all halls, and lessons learned along the way.

        Speaker: Thomas Britton (JLAB)
      • 12
        Fast 2D Bicephalous Convolutional Autoencoder for Compressing 3D Time Projection Chamber Data
        Speaker: Yi Huang (Brookhaven national lab)
      • 13
        Autonomous selection of physics events: A RHIC demonstrator for EIC physics

        With ever-increasing collision rates producing larger data volumes and the search for rarer physics processes, it is becoming apparent that autonomic decision making can play a key role in tagging physics events of interest to specific groups or filtering data streams to manageable levels. A demonstrator for separating events with a heavy flavor decay from background events in proton-proton collisions at the Relativistic Heavy Ion Collider (RHIC) with the sPHENIX detector is presented. Due to data volume limitations, sPHENIX is capable of recording 10% of the minimum-bias collisions at RHIC using streaming readout in addition to its 15 kHz hardware trigger of rare events. This demonstrator will use machine-learning algorithms on FPGAs to sample the remaining 90% of the collisions, determine the event topology and send a decision to the data acquisition system to record events of interest. The design of the demonstrator and its use at sPHENIX will be refined for deployment at the EIC for the tagging of heavy flavor decays and DIS-electrons.

        Speaker: Cameron Dean (MIT)
      • 14
        Discussion
    • Tutorial 1
      • 15
        Tutorial: Continual Learning
        Speaker: Andrea Cossu (University of Pisa)
    • AI/ML for Accelerators
      Conveners: Elena Fol (member@cern.ch), Kevin Brown (C-AD)
      • 16
        Bayesian Optimization Techniques for Accelerator Control and Characterization

        Future improvements in accelerator performance are predicated on increasing capabilities in online control of beams inside accelerators. Machine learning techniques have been the focus of work at SLAC to increase our ability to autonomously optimize and characterize beam dynamics inside accelerator facilities. Bayesian optimization algorithms, which leverage statistical surrogate models of objective functions to effectively address complex optimization challenges, are well situated for solving online optimization challenges in accelerator science. We describe Bayesian optimization techniques that have been developed to solve a wide range of online accelerator control problems, including single and multi-objective optimization, autonomous characterization, with or without constraints, high dimensional parameter spaces, and in the presence of hysteresis effects. These techniques can be used to effectively automate routine accelerator facility processes and enable novel capabilities in current and future accelerator systems.

        Speaker: Ryan Roussel (SLAC National Laboratory)
      • 17
        Machine learning for digital twin development and polarization optimization at BNL hadron injectors

        Polarization optimization collaboration (BNL, Cornell, JLab, SLAC, RPI)

        Speaker: Lucy Lin (Cornell University)
      • 18
        Beam Condition Forecasting with non-destructive measurements at FACET-II

        Beam diagnostic technology is one of the foundations of large particle accelerator facilities. A challenge with operating these systems is the measurement of beam dynamics. Many methods such as beam position monitors have an inherent destructive quality to the beam and produce perturbations after the measurement. The ability to measure the beam conditions with non-destructive edge radiation allows for us to have a more stable understanding and predictability of the beam condition. We are developing a machine learning workflow for the downstream prediction and future forecasting of the beam condition utilizing the non-destructive edge radiation measurements and novel graph neural networks in collaboration with FACET-II at SLAC. Our methods divide the problem into two different aspects. First, we are developing machine learning algorithms with the beam physics integrated within each layer of the network. Second, developing an online surrogate model of edge radiation using SRW will allow for automatic generation of new beam states due to the changing parameters of accelerator facilities over time. We plan to integrate and test our prediction system at the SLAC facility to perform beam condition prediction and verification at FACET-II.

        Speaker: Matt Kilpatrick (RadiaSoft)
      • 19
        Machine Learning applications for collider luminosity maximization

        The luminosity of a collider can be affected by many parameters at the same time. It is not easy to distinguish the effects of one parameter from all other parameters separately. Therefore, optimizing the performance of a collider such as RHIC, EIC becomes a multi-objective optimization problem with possible noisy signals and involves many parameters. Therefore, machine learning (Bayesian and Gaussian Process) could be a good tool for the luminosity optimization. Here, we talk about a Bayesian optimization method which is developed at LBNL GPTune and its planned application to RHIC luminosity optimization, as well as its possible application to EIC collider.

        Speakers: Ji Qiang (LBNL), Sherry Li (LBNL), Will Fung (MSU), Xiaofeng Gu (Collider Accelerator Department, BNL), Yi-Kai Kan (LBNL), Yue Hao (Michigan State University / Brookhaven National Laboratory)
      • 11:20
        Coffee Break
      • 20
        Anomaly detection at an X-ray FEL

        Modern light sources produce too many signals for a small operations team to monitor in real time. As a result, recovering from faults can require long downtimes, or even worse subtle performance issues may persist undiscovered. Existing automated methods tend to rely on pre-set limits which either miss subtle problems or produce too many false positives. AI methods can solve both problems, but deep learning techniques typically require extensive labeled training sets, which may not exist for anomaly detection tasks. Here we will show work on unsupervised AI methods developed to find problems at the Linac Coherent Light Source (LCLS). Whereas most unsupervised AI methods are based on distance or density metrics, we will describe a coincidence-based method that identifies faults through simultaneous changes in sub-system and beam behavior. We have applied the method to radio-frequency (RF) stations faults — the most common cause of lost beam at LCLS — and find that the proposed method can be fully automated while identifying 50% more events with 6x fewer false positives than the existing alarm system. We will also show work on a general outlier detection method, including an example of finding a previously unknown beam-scraping event.

        Speaker: Daniel Ratner (staff@stanford.edu;member@stanford.edu)
      • 21
        Uncertainty estimation and RL applications at JLab

        Standard deep learning models for classification and regression applications are ideal for capturing complex system dynamics.
        Unfortunately, their predictions can be arbitrarily inaccurate when the input samples are not similar to the training data. Implementation of distance aware uncertainty estimation can be used to detect these scenarios and provide a level of confidence associated with their predictions.
        We present results using UQ for ML methods for 1) anomaly detection at Spallation Neutron Source (SNS) accelerator and 2) the Fermi National Accelerator Lab (FNAL) Booster Accelerator Complex.
        We also present an application of reinforcement learning to improve accelerator controls.

        Speaker: Malachi Schram (Thomas Jefferson National Accelerator Facility)
      • 22
        Using Machine Learning to Improve Dynamic Aperture Estimates

        The dynamic aperture (DA) is an important concept in the study of nonlinear beam dynamics. Several analytical models used to describe the evolution of DA as a function of time, and to extrapolate to realistic time scales that would not be reachable otherwise due to computational limitations, have been successfully developed. Even though these models have been quite successful in the past, the fitting procedure is rather sensitive to several details. Machine Learning (ML) techniques carry the potential to address some of these challenges. Two applications of ML approaches will be presented and discussed in detail. Firstly, ML has been used to efficiently detect outliers in the DA computations. Secondly, ML techniques have been applied to improve the fitting procedures of the DA models, thus improving their predictive power

        Speaker: Frederik Van der Veken (CERN)
      • 23
        Discussion
    • 13:00
      Lunch
    • AI/ML for Data Analysis and Theory
      Conveners: Brandon Kriesten (ANL), Vinicius Mikuni (Other@lbl.gov;Staff@lbl.gov;Member@lbl.gov)
      • 24
        The EXCLAIM collaboration approach to deeply virtual exclusive processes
        Speaker: simonetta liuti (university of virginia)
      • 25
        Decoding inverse problems in QCD with ML-based algorithms
        Speaker: Yaohang Li (Old Dominion University)
      • 26
        What can AI do for lattice-QCD parton distribution calculations?
        Speaker: Huey Wen (MSU)
      • 15:15
        Coffee Break
      • 27
        Interpretable Machine Learning applications to Jet Background Subtraction

        Reconstructing jets in heavy collisions has always required dealing with the challenges of a high background environment. Traditional techniques, such as the area based method, suffered from poor resolution at low momenta due to the large fluctuating background there. In recent years, the resolution has been improved by using machine learning to estimate the background. While machine learning tends to lead to improvements in general (wherever it is applied), care must be taken to ensure these improvements do not come at the cost of interpretability or bias from models used for training. We demonstrate a middle path – using machine learning techniques to translate “black-box” models (such as neural nets) into human interpretable formulas. We present a novel application of symbolic regression to extract a functional representation of a deep neural network trained to subtract background for measurements of jets in heavy ion collisions. With this functional representation we show that the relationship learned by a neural network is approximately the same as a new background subtraction method using the particle multiplicity in a jet. We compare the multiplicity method to the deep neural network method alone, showing its increased interpretability and comparable performance. We also discuss the application of these techniques to background subtraction for jets measured at the EIC.

        Speaker: Charles Hughes (Iowa State University)
      • 28
        High Dimensional Unfolding using Machine Learning
        Speaker: Benjamin Nachman
      • 29
        Hadronization Models Using Machine Learning
        Speaker: Manuel Szewc (University of Cincinnati)
      • 30
        Discussion
    • Tutorial 2
    • Foundation Models and Trends in Data Science with Tutorials
      Conveners: Daniel Murnane (LBL), Yaohang Li (Old Dominion University)
      • 32
        Introduction and Overview of Foundation Models
        Speaker: Kazuhiro Terao (SLAC National Accelerator Laboratory)
      • 33
        The Trillion Parameter Consortium
        Speaker: Rick Stevens (ANL)
      • 34
        A Large Language Model-based Assistant for the Electron Ion Collider

        A large-scale physics experiment is distinguished by a wide collaboration and array of research topics, and frequently encounters obstacles in information exchange coordination and data analysis. We introduce an assistant model designed to enhance collaboration and streamline data analysis. This model acts as an intelligent intermediary for research teams, enabling information exchange, offering expert insights, and supporting decision-making for researchers. Moreover, it can process and interpret experimental data in real-time, facilitating quicker researcher decisions. In this work, we focus on the Electron Ion Collider (EIC), the next generation facility for Quantum Chromodynamics. The EIC has already engaged hundreds of collaborators, producing a wealth of research outputs. Significantly, it has integrated artificial intelligence and machine learning into various facets of its scientific endeavors. Using Large Language Models (LLMs) for summary generation and tailored prompt engineering, we introduce a framework for the EIC. We showcase proof-of-concept studies, efficiently curating information from a vector database containing technical notes and other EIC community contributions. Demonstrating the viability of our method, we present a web application that visualizes the outcomes from our proof-of-concept studies. We conclude with insights for potential advancements and discussions on future endeavors.

        Speaker: Dr Karthik Suresh (William & Mary )
      • 35
        Chatlas - the ATLAS AI Assistant
        Speaker: Daniel Murnane (LBL)
      • 11:30
        Coffee break
      • 36
        TrackBERT - Generalist Learning of Trackers for Downstream Tasks
        Speaker: Xiangyang Ju (Lawrence Berkeley National Laboratory)
      • 37
        Adversarial Methods for Generative Data Understanding
        Speaker: Eric Yeats (Duke)
      • 38
        Toward a generative modeling analysis of CLAS exclusive 2π photoproduction
        Speaker: Tareq Alghamdi (ODU)
      • 39
        Discussion
    • 13:00
      Lunch
    • AI/ML in Production, Distributed ML
      Conveners: David Lawrence (Jefferson Lab), Rui Zhang (University of Wisconsin-Madison)
      • 40
        Machine learning as a service for HEP using cloud
        Speaker: Luca Giommi (University of Bologna)
      • 41
        Particle identification with machine learning from incomplete data in the ALICE experiment
        Speaker: Maja Karwowska (CERN, Warsaw U. of Tech.)
      • 42
        Scalable AI/ML Workflow Management Across Distributed Heterogeneous Resources With PanDA

        The Production and Distributed Analysis (PanDA) system originating in LHC's ATLAS experiment has been steadily evolving upon a technical foundation of proven scalability and extensibility, extending in recent years to new experiments (Rubin Observatory, sPHENIX) and new capabilities in managing large scale, complex workflows across diverse geographically distributed resources. AI/ML has been a particular focus, with support for processing intensive workflows benefiting from extensive automation and access to large scale resources, such as hyperparameter optimization (HPO). This presentation will introduce PanDA, its complex workflow management capabilities and their practical application in HPO and other AI/ML workflows. Finally, PanDA's role in a new R&D program providing a scalable and distributed workflow engine for AI-assisted EIC detector design will be described.

        Speaker: Wen Guan (BNL)
      • 43
        The Exa.TrkX Project
        Speaker: Xiangyang Ju (Lawrence Berkeley National Laboratory)
      • 15:30
        Coffee break
      • 44
        ATLAS data analysis using a parallelized workflow on distributed cloud-based services with GPUs
        Speaker: Jay Sandersara (CERN)
      • 45
        Online Track Reconstruction and Physics Analysis in CLAS12
        Speaker: Gagik Gavalian (Jefferson Lab)
      • 46
        MLaaS4HEP, Machine Learning for Columnar High Energy Physics Analysis
        Speaker: Elliott Kaufman (Princeton University)
      • 47
        Discussion
    • Closing Caldwell auditorium

      Caldwell auditorium

      Catholic University of America, Washington D.C.

    • Tutorial 3 Caldwell auditorium

      Caldwell auditorium

      Catholic University of America, Washington D.C.

    • Hackathon Caldwell auditorium

      Caldwell auditorium

      Catholic University of America, Washington D.C.