14 October 2024
Convergence Center @ Purdue University
US/Eastern timezone

Contribution List

48 out of 48 displayed
Export to PDF
  1. Seungbin Park, Yulei Zhang (University of Washington (US))
    14/10/2024, 09:00
  2. Dr Song Han (MIT)
    14/10/2024, 09:10

    This talk presents efficient multi-modal LLM innovations with algorithm and system co-design. I’ll first present VILA, a visual language model deployable on the edge. It is capable of visual in-context learning, multi-image reasoning, video captioning and video QA. Followed by SmoothQuant and AWQ for LLM quantization, which enables VILA deployable on edge devices, bringing new capabilities for...

    Go to contribution page
  3. Pan Li
    14/10/2024, 09:35

    Graphs have been widely applied to model intricate relationships among entities. The application of Graph Machine Learning (GML) to enhance prediction capabilities for graph-structured data is prevalent in several scientific disciplines, such as particle physics, material science, and biology. However, applications in these domains often present challenges due to changes in data distributions...

    Go to contribution page
  4. Will Benoit
    14/10/2024, 10:00

    Multi-messenger astronomy is one of the pillars of A3D3. It spans optical, neutrino, and gravitational wave astronomy, each of which is a field with exciting physics and the potential to apply advanced machine learning techniques. In this presentation, I will give an overview of the work that is currently being done across all of the groups in this area.

    Go to contribution page
  5. Seungbin Park
    14/10/2024, 10:15
  6. Yuan-Tang Chou (University of Washington (US))
    14/10/2024, 10:30
  7. Siqi Miao
    14/10/2024, 10:45

    Algorithm and Hardware Co-Development (HAC) is a key focus area within A3D3, supporting the institute’s mission to build accelerated AI solutions for scientific discovery. Our team develops AI algorithms to address significant challenges in data-driven research, including data irregularity, label scarcity, and the complexity of understanding AI models, while also providing efficient hardware...

    Go to contribution page
  8. Advaith Anand (University of Washington (US))
    14/10/2024, 11:00

    Harnessing the Data Revolution (HDR), is an effort by the National Science Foundation (NSF) to promote the exploration of fundamental scientific questions using data-driven techniques. To raise interest in these approaches, and the HDR community, we have developed a Machine Learning (ML) challenge for anomaly detection, taking advantage of widespread data from several HDR institutes. This...

    Go to contribution page
  9. Zhijian Liu (Massachusetts Institute of Technology)
    14/10/2024, 11:45
  10. Julia Lynne Gonski (SLAC National Accelerator Laboratory (US))
    14/10/2024, 11:50
  11. Dylan Sheldon Rankin (University of Pennsylvania (US))
    14/10/2024, 11:55
  12. Dylan Sheldon Rankin (University of Pennsylvania (US)), Julia Lynne Gonski (SLAC National Accelerator Laboratory (US)), Zhijian Liu (Massachusetts Institute of Technology)
    14/10/2024, 12:00
  13. Kira Nolan
    14/10/2024, 13:45

    In astronomy, the successful identification of electromagnetic counterparts to gravitational wave signals unlocks unique science that is otherwise impossible with siloed observations. Efforts in multi-messenger astronomy stand to become increasingly fruitful
    but also more complex over the next decade as new instruments provide exponentially larger data streams. Not only does this work...

    Go to contribution page
  14. Rian Flynn (Purdue University (US))
    14/10/2024, 13:55

    The High-Luminosity Large Hadron Collider (HL-LHC), anticipated to begin operations in 2029, will generate data at an astounding rate on the order of 100 terabits per second. To efficiently process and filter these data, the Compact Muon Solenoid (CMS) experiment
    relies on the extremely low-latency Level-1 trigger, which uses Field-Programmable Gate Arrays (FPGAs). My project focuses on...

    Go to contribution page
  15. Alexandra Junell Brown
    14/10/2024, 14:05

    The Zwicky Transient Facility is capable of triggering hundreds of thousands of alerts a night for potential astronomical transients. Quickly classifying these objects is critical for determining candidates for follow up observations. Machine learning already plays a role in the transient classification pipeline, and has shown success training on image time series and photometric time series....

    Go to contribution page
  16. Miles Cochran-Branson (University of Washington (US))
    14/10/2024, 14:15

    Particle tracking at Large Hadron Collider (LHC) experiments is a crucial component of particle reconstruction, yet it remains one of the most computationally challenging tasks in this process. As we approach the High-Luminosity LHC era, the complexity of tracking is expected to increase significantly. Leveraging coprocessors such as GPUs presents a promising solution to the rising...

    Go to contribution page
  17. Janina Dorin Hakenmueller (Duke University)
    14/10/2024, 14:28

    Core collapse supernova explosions offer a rich potential of physics to explore. The emitted neutrinos are the first signals to reach the earth. Detecting these neutrinos and their direction can provide valuable information to optical detection systems in a multi messenger astronomy approach.
    In liquid argon time projection chambers such as DUNE the charge interactions are the most abundant...

    Go to contribution page
  18. Josh Peterson
    14/10/2024, 14:41

    IceCube DeepCore is an infill of the IceCube Neutrino Observatory designed to study neutrinos with energies as low as 5 GeV. Reconstruction and classification tasks near the lower energy threshold of IceCube DeepCore are especially difficult due to the low number of detected photons per neutrino event. Many neural networks have been developed for these tasks, and there are many ways we could...

    Go to contribution page
  19. Melissa Quinnan (Univ. of California San Diego (US))
    14/10/2024, 14:54
  20. Jason Weitz (UCSD)
    14/10/2024, 15:30

    We develop an automated pipeline to streamline neural architecture codesign for physics applications, to reduce the need for ML expertise when designing models for a novel task. Our method employs a two-stage neural architecture search (NAS) design to enhance these models, including hardware costs, leading to the discovery of more hardware-efficient neural architectures. The global search...

    Go to contribution page
  21. ChiJui Chen
    14/10/2024, 15:43

    In software-hardware co-design, balancing performance with hardware constraints is critical, especially when using FPGAs for real-time applications in scientific fields with hls4ml. Limited resources and stringent latency requirements exacerbate this challenge. Existing frameworks such as AutoQKeras use Bayesian optimization to balance model size/energy, and accuracy, but they are...

    Go to contribution page
  22. Alexander Yue
    14/10/2024, 15:56

    Detectors at next-generation high-energy physics experiments face several daunting requirements: high data rates, damaging radiation exposure, and stringent constraints on power, space, and latency. To address these challenges, machine learning (ML) in readout
    electronics can be leveraged for smart detector designs, enabling intelligent inference and data reduction at-source. Autoencoders...

    Go to contribution page
  23. Rajeev Botadra
    14/10/2024, 16:09

    Non-Human Primates (NHPs) are central to neuroscience research due to their complex behavioral interactions and physiological similarities to the human brain. A principal motivation behind the NHP research in the aoLab at the University of Washington is to understand and model neural circuits, which can be translated for practical applications for humans. However, the nonlinear...

    Go to contribution page
  24. Seungbin Park
    14/10/2024, 17:15

    Decoding neural activity into behaviorally-relevant variables such as speech or movement is an essential step in the development of brain-machine interfaces (BMIs)and can be used to clarify the role of distinct brain areas in relation to behavior. Two-photon (2p) calcium imaging provides access to thousands of neurons withsingle-cell resolution in genetically-defined populations and therefore...

    Go to contribution page
  25. Seiya Tsukamoto
    14/10/2024, 17:20

    The detection of gravitational waves with the Laser Interferometer Gravitational Wave Observatory (LIGO) has provided the tools to probe the furthest reaches of the universe. A rapid follow up to compact binary coalescence (CBC) events and their electromagnetic counterparts is crucial to find short lived transients. After a gravitational wave (GW) detection, another particular challenge is...

    Go to contribution page
  26. Derrick Appiah Osei
    14/10/2024, 17:25

    Field-Programmable Gate Arrays (FPGAs) are increasingly becoming pivotal in the advancement of artificial intelligence (AI) and deep learning applications. Their unique architecture allows for customizable hardware acceleration, which is instrumental in handling the intensive computational demands of modern AI algorithms.

    Transmission Electron Microscopy (TEM) provides exceptional...

    Go to contribution page
  27. Megan Averill
    14/10/2024, 17:30

    Multi-Messenger observations of kilonovae can be used to measure the Hubble Constant by combining distance posteriors from the gravitational wave observations with a redshift measurement of the source’s host galaxy. There is a significant discrepancy between two existing, prominent estimates of the Hubble constant: Planck, utilizing cosmic microwave background radiation and lambda cold dark...

    Go to contribution page
  28. Mariel Peczak
    14/10/2024, 17:35

    The upcoming high-luminosity upgrade to the LHC will involve a dramatic increase in the number of simultaneous collisions delivered to the Compact Muon Solenoid (CMS) experiment. To deal with the increased number of simultaneous interactions per bunch crossing as well as the radiation damage to the current crystal ECAL endcaps, a radiation-hard high-granularity calorimeter (HGCAL) will be...

    Go to contribution page
  29. Kira Nolan (California Institute of Technology)
    14/10/2024, 17:40

    In this work we show advancements in follow-up methods for detection of electromagnetic counterparts to gravitational wave signals. These multi-messenger observations are important targets for their ability to unlock science including measurement of the Hubble constant, which is a current major effort in cosmology. In this work we include a data-driven heuristic to select anomalous flares...

    Go to contribution page
  30. Emma de Bruin
    14/10/2024, 17:45

    Aframe is a gravitational wave search pipeline being constructed at the University of Minnesota Twin Cities. It is a machine learning pipeline designed to look for gravitational wave signals. It performs well for events with a high chirp mass, but could be better for ones with a lower chirp mass. My work is on improving Aframe’s performance for these lower mass events using a new type of Q...

    Go to contribution page
  31. Julian Goddy
    14/10/2024, 17:50

    Deploying lightweight models on FPGAs requires robust workflows for tracking, saving, and transferring model information, and ensuring that this information adheres to FAIR (Findable, Accessible, Interoperable, and Reproducible) principles. We present a Python package that automates the identification and documentation of key metadata for machine learning models developed in PyTorch or...

    Go to contribution page
  32. Mostafa Cham (iHARP)
    14/10/2024, 17:55

    Accurate estimation of subglacial bed topography is crucial for understanding ice sheet dynamics and their responses to climate change. In this study, we employ machine learning models, enhanced with Spark parallelization, to predict subglacial bed elevation using surface attributes such as ice thickness, flow velocity, and surface elevation. Radar track data serves as ground truth for model...

    Go to contribution page
  33. Alexandra Junell Brown
    14/10/2024, 18:00

    In time-domain astronomy, rapid classification of astronomical transients is critical for determining candidates for follow-up observations. With the advent of the Vera Rubin Observatory’s Legacy Survey of Space and Time, the backlog of astronomical data will increase by terabytes a night. Machine learning models capable of processing and analyzing large quantities of data can advance the...

    Go to contribution page
  34. Steven Henderson (UMN - Twin Cities)
    14/10/2024, 18:05

    Binary black hole mergers can be located by collecting and analyzing the unique gravitational wave signals they produce. Deep learning computational models, specifically Aframe, are used to identify and filter gravitational wave signals more accurately and in less time than traditional matched filtering analyses. The current machine learning model that we use, Aframe, was originally developed...

    Go to contribution page
  35. Jared Burleson (University of Illinois at Urbana-Champaign)
    14/10/2024, 18:10

    The next phase of high energy particle physics research at CERN will
    involve the High-Luminosity Large Hadron Collider (HL-LHC). In preparation for
    this phase, the ATLAS Trigger and Data AcQuisition (TDAQ) system will undergo
    upgrades to the online software tracking capabilities. Studies are underway to
    assess a heterogeneous computing farm deploying GPUs and/or FPGAs, together
    with the...

    Go to contribution page
  36. David Jiang (Univ. Illinois at Urbana Champaign (US))
    14/10/2024, 18:15

    Pixel detectors are highly valuable for their precise measurement of charged particle trajectories. However, next-generation detectors will demand even smaller pixel sizes, resulting in extremely high data rates surpassing those at the HL-LHC. This necessitates a “smart” approach for processing incoming data, significantly reducing the data volume for a detector’s trigger system to select...

    Go to contribution page
  37. Cymberly Tsai
    14/10/2024, 18:28

    Fitting data to a variety of models is a fundamental challenge in the monitoring and control of dynamical systems across science and manufacturing domains. In this work, we present a compact foundation model designed for adaptive function selection and regression. The proposed architecture utilizes 1D convolutional neural networks (CNNs), augmented by physical constraints, to facilitate the...

    Go to contribution page
  38. Dr Song Han (MIT)
  39. Pan Li
  40. Melissa Quinnan (Univ. of California San Diego (US))
  41. Siqi Miao
  42. Will Benoit