By: Yuan-Tang Chou

Grace Williams

April 2, 2024

The 22nd International Workshop on Advanced Computing and Analysis Techniques in Physics Research (ACAT 2024) is an exciting workshop held every 18 months. It brings together computational experts from around the world who work in diverse fields of physics, like particle and astrophysics, to discuss advanced technology with the potential to enhance research in physics.. 

This year’s theme is “Foundation Models for Physics—Nexus of Computation and Physics through Embracing the Era of Foundation Models.” Our researchers, Vladimir Loncar and Javier Duarte, chaired two of the exciting sessions. 

Below, we highlight some of the many of the exciting contributions from A3D3 members that were presented during the week-long workshop.


On Monday afternoon, Zihan Zhao (A3D3 Grad student) presented his work on Self-supervised Learning for Jet Tagging,”  demonstrating the potential of self-supervised learning to utilize real, unlabeled data for more efficient and accurate jet classification. 

Zihan Zhao presenting at ACAT 2024


Researcher Javier Duarte and steering board member Nhan Tran (Fermilab Wilson Fellow) chaired Tracks Two: Data Analysis – Algorithms and Tools which discussed advanced ML algorithms for physics object reconstruction in the particle experiment. 

Nhan also gave a plenary on Tuesday about “AI and Microelectronics for Science.”

Wahid Bhimji, A3D3 steering board, presented “Fair Universe: HiggsML Uncertainty Challenge” which brings FAIRness to the ML community, and also convened the plenary on Thursday. 

Pic: Wahid Bhimji Convened plenary 

In the afternoon session at the “Towards the Construction of Foundational Models at the LHC” talk given by Phil Harris (A3D3 deputy director), he presented recent work that proposed novel Self-Supervised Learning Strategies. The method constructs a space that preserves discrimination power and reduces the impact of systematic uncertainties. 

Xiangyang Ju (LBNL computing system engineer, A3D3 affiliated member) talked about a novel idea to use a large language model to perform particle tracking in his “Leveraging Language Models for Particle Reconstruction.”   

Xiangyang Ju presents on the use of large language models in particle tracking


On Thursday morning, Aobo Li (UCSD Professor) gave a plenary talk about “Detecting Rare Events Using Artificial Intelligence.” He proposed a unified multimodal foundation model for all rare event search experiments.

 “This could forge the experiments like a union,” Aobo said.

Aobo Li gives his presentation on “Detecting Rare Events Using Artificial Intelligence”

In the afternoon, Yuan-Tang Chou (A3D3 Postdoc) presented “ACTS as a Service” and demonstrated how to utilize heterogeneous computing resources better to accelerate track reconstruction using the Triton GPU Inferences server.

“We’re proposing this idea to better utilized heterogeneous computing systems to accelerate the track reconstructions in ACTS. I think people are quite interested in how this can actually apply in the current production framework.”

-Yuan-Tang Chou

By: Deep Chatterjee

December 27, 2023

New Orleans, LA – The 37th Conference on Neural Information Processing Systems was held in New Orleans between Dec 10 – 16, 2023. The Machine Learning and the Physical Sciences Workshop at NeurIPS brought together researchers pursuing applications of Machine Learning techniques in Physics, and developing new techniques based on physical concepts like conversation laws and symmetries. There was a total of 250 accepted papers for the workshop poster session – both in-person and remote. While most papers were related to Physics and Astronomy, there were several papers on applications in medical sciences, material science, and earth sciences. A3D3 had a prominent presence at the workshop, with a contributed talk, 4 papers from A3D3 members, and 2 papers from A3D3 Steering Board committee members’ team. 

Elham E Khoda from UW Seattle presented on the implementation of Transformers on FPGA using HLS4ML, for low-latency applications like L1 triggers at the LHC and ATLAS, and searches for anomalous signals in gravitational-wave data. 

Elham Khoda presents a contributed talk on the implementation of transformers in HLS4ML

Deep Chatterjee from MIT presented a poster on optimizing likelihood-free inference by marginalizing nuisance parameters using self-supervision. [paper link

Niharika Sravan from Drexel University presented Pythia – a reinforcement learning model that maximizes the search for kilonovae in the presence of several contaminant objects from the Zwicky Transient Facility. [paper link

Anni Li from UCSD gave a poster on Induced Generative Adversarial Particle Transformers on the use of induced particle-attention blocks to surpass existing particle simulation generative models. [paper link

Deep Chatterjee (left) and Alex Gagliano (right) present posters.

There were two posters presented from A3D3 steering board members. Ashley Villar (along with Alex Gagliano) gave a poster on convolutional variational autoencoder to estimate the redshift, stellar mass, and star-formation rates of galaxies from multi-band imaging data. [paper link] Nhan Tran (along with C. Xu) have a poster on a Proximal Policy Optimization (PPO) algorithm to uniform proton beam intensity for the Mu2e experiment at Fermilab. [paper link

By: Katya Govorkova and Yuan-Tang Chou
November 27, 2023

The “AI and the Uncertainty Challenge in Fundamental Physics” workshop was a dynamic experience filled with experts from different areas of applied AI in science. Experts from fundamental science, computer science, and statistics exchange ideas on how to incorporate uncertainty in AI. The workshop took place at the Sorbonne Center for Artificial Intelligence (SCAI) in Paris and at Institut Pascal Université Paris-Saclay in Orsay, France.

Researcher Mark Neubauer (Faculty at the University of Illinois at Urbana-Champaign, A3D3 co-PI) led the discussion in the Uncertainty Quantification session on Tuesday and emphasized the idea of explainable AI.

The session on Wednesday afternoon was specifically dedicated to the FAIR Universe ML challenge, which addresses uncertainties in physics through AI. A3D3, represented by Yuan-Tang Chou (A3D3 Postdoc at the University of Washington) and Katya Govorkova (A3D3 Postdoc at the Massachusetts Institute of Technology), shared valuable insights from organizing an anomaly detection data challenge, emphasizing the necessity of robust frameworks and collaboration. Lessons learned from Katya’s experiences underscored the crucial role challenges play in advancing the field. She also underlined what can be improved in future challenges. 

Katya Govorkova presenting Exploring Data Challenges and Leveraging Codabench: A Practical Journey with unsupervised New Physics detection at 40 MHz.

Yuan-Tang highlights another ML challenge example for NSF HDR institute beyond particle physics. The challenge utilized the Neuroscience dataset provided by Prof. Dadarlat’s lab ( Purdue University, A3D3), which tried to decode limb trajectories from neuro activity with ML.

Yuan-Tang presenting ML challenge using Neuroscience dataset

The Codabench platform, noted for its versatility, was highlighted as instrumental in organizing public challenges. The public showed a great interest in hosting and participating in challenges. Challenges were recognized as bridges connecting different disciplines, particularly physics and computer science. The event highlighted those challenges.

The participants provided many great ideas and suggestions during the one-week workshop. Elham E Khoda (Postdoc at the University of Washington, A3D3) gave an excellent summary on the last day to discuss the lesson learned during the week and the next step to improve the HiggsML Uncertainty Challenge. “We definitely want not only people outside the particle physics to join the ML Challenge,” Elham said. “We also encourage participation outside of the domain who can think differently and come up with innovative ideas.”

By: Rajeev Bhavin Botadra

November 22, 2023

Researchers Javier M. Duarte (Faculty at the University of California San Diego, A3D3), Luke Song (Graduate student at the Ohio State University, Imageonics), and Rajeev B. Botadra (Graduate student at University of Washington, A3D3) represented the National Science Foundation (NSF) Harnessing the Data Revolution (HDR) programs at the 2023 National Diversity in STEM (NDiSTEM) conference hosted in Portland, Oregon.

The SACNAS NDiSTEM conference is held annually to provide a platform for underrepresented groups in STEM to connect with peers and mentors and explore opportunities within academia and industry. By participating in this event the team aims to promote the programs under the HDR initiative while furthering the broader NSF effort of diversity in STEM.

The representatives presented opportunities across all five institutions under the HDR grant,  emphasizing the different scientific applications studied under each branch as they aligned with students’ interests. They also shared their experiences and career journeys, advising students unsure of the next step in their careers and making connections for future collaborations.

“The SACNAS NDiSTEM Conference is by far the largest gathering of its kind in the country,” Prof. Duarte said. “It’s a unique opportunity to reach potential trainees that we may not find at other conferences. Everyone is very open about sharing their cultures and identities because they recognize that it’s not separate from their science.”

The team set up a simple Pokémon classification demo at the booth using a webcam, a Pynq-Z2 FPGA board, and a monitor for display output. Using a simple quantized ResNet model fine-tuned on an open-source pokemon dataset, the Pynq-Z2 classified Pokémon in front of the camera and transmitted the labeled output to the external monitor. Rajeev commented, “the demo was very helpful in drawing people’s attention amongst dozens of other booths and breaking the ice towards a longer conversation about our research.”

“We got to interact with so many people and they were so excited to find out about the HDR  research opportunities,” Prof. Duarte said. “We hope to come back every year!”

By: Xiangyang Ju (Computing System Engineer, LBNL)

November 13, 2023

In the vibrant Mile High City, members and affiliates of the A3D3 institute converged with a gathering of over 100 scientists, engineers, and educators. Their mission: to fortify the Harnessing the Data Revolution (HDR) ecosystem. This annual HDR-wide conference, now in its second year, aims not only to strengthen the HDR ecosystem but also to extend its reach to other related NSF-supported initiatives, fostering collaboration in our collective endeavors.

The HDR initiative, funded by the National Science Foundation (NSF), is a nationwide effort that commenced in 2016. It seeks to enable new avenues of data-driven discovery, addressing fundamental questions at the forefront of science and engineering. Within the HDR ecosystem, the A3D3 institute spearheads a paradigm shift by deploying real-time artificial intelligence on a grand scale to advance scientific knowledge and expedite the process of discovery.

The conference had four primary objectives for its participants:

  1. Foster community-building, forging stronger ties among HDR entities and the broader data-intensive research communities.
  2. Facilitate cross-learning by building on successes, best practices, and innovative products.
  3. Provide a platform for reflection on the achievements and future goals of each HDR entity.
  4. Identify overarching challenges in data-intensive research, not only among HDR entities but also beyond. The conference aimed to foster new collaborations and explore future opportunities.

To meet these goals, the conference featured a diverse range of activities, including keynote presentations, discussion panels, and “unconference” sessions. Notable activities included a pitch session and subsequent asset mapping for the winning pitches, in which the A3D3 Institute made substantial contributions.

In summary, A3D3 played a pivotal role in the conference, leading the reflection session titled “Humans in the Equation: Stories of Collaboration.” The session featured a success story narrated by Javier Duarte from UC San Diego, highlighting the Postbaccalaureate program organized by the A3D institute. This program aims to enhance access to scientific careers for post-baccalaureates interested in gaining insights into scientific research and exploring various scientific domains. Professor Duarte’s presentation was followed by cross-community discussions on “Equitable, Diverse, and Inclusive Training, Education, and Outreach Opportunities in the HDR Ecosystem,” co-led by Professor Mark Neubauer.

During the same session, A3D3 members engaged in developmental reflections, examining the context of the ecocycle planning phases and potential obstacles. The primary purpose of this activity was to identify impediments and opportunities for progress.

The six posters given by A3D3 members covered a wide range of scientific frontiers. 

  • Brian Healy: the “Machine Learning Classification of Time-varying Astrophysical Sources” 
  • Seungbin Park: “Decoding multi-limb running trajectories from two-photon calcium imaging using deep learning.”
  • Ben Carlson: “Module for evaluation of machine learning algorithms in FPGA hardware for high energy physics.”
  • Mark Neubauer: “A3D3 community engagement, education and outreach”
  • Javier Duarte: “hls4ml: Open-source codesign of machine learning algorithms on FPGA for scientific discovery.”
  • Phil Harris: “Real-time Gravitational Wave Alerts using AI.”

Figures: left: Brian Healy. Middle: Seungbin Park, Right: Ben Carlson

A3D3 took the lead in the topical session addressing equity, diversity, and inclusion (EDI) within the realm of training, education, and outreach. The session served as a platform for discussing the promotion of EDI in workforce development within the HDR ecosystem. Key topics of discussion included ensuring fair trainee selection processes, establishing, maintaining, and enhancing specialized training programs, creating specific forums, and extending program impact through education and outreach. During the session, A3D3 members shared their best practices with other institutes within the HDR ecosystem, covering areas such as equitable trainee recruitment, outreach event organization, and the establishment of post-baccalaureate programs.

A separate session led by A3D3 delved into the realm of machine learning challenges. Professor Philip Harris, representing MIT, delivered a comprehensive presentation titled “Machine Learning Challenges, FAIR, and Reproducible Machine Learning Workflows.” In this talk, Philip introduced the latest technical developments in the Codabench platform, designed to host diverse challenges and facilitate automated metric calculations. Many other HDR institutes expressed keen interest in using this platform for publishing machine learning challenges. The discussions extended into an ideation exploration session, during which Philip Harris advocated for the inception of an Anomaly Detection challenge spanning all HDR institutes—a pan-HDR “Grand Challenge.”

As the annual HDR conference concluded, it left attendees inspired by new ideas and fresh possibilities. Looking ahead, A3D3 will take the helm next year, hosting the conference at UIUC in Urbana-Champaign.

By: Patrick McCormack (Postdoc MIT, A3D3)

October 30, 2023

At the workshop, seven A3D3 trainees gave presentations on their work.  Leading off, Farouk Mokhtar of UCSD and Santosh Parajuli of UIUC presented their work on implementing machine learning models for the LHC experiments CMS and ATLAS, respectively.  Though working on independent efforts, both use graph neural networks to efficiently and scalably reconstruct particles. Alongside the first smatterings of autumn leaves, an international assortment of more than 150 Physicists, Computer Scientists, Engineers (and more) descended upon Imperial College London (ICL) this past week.  There they enjoyed the crisp weather and the fourth iteration of the Fast Machine Learning for Science (FastML) Workshop, which ran from September 25-28.

The FastML workshop series was born in 2019 as a small and informal workshop focused on High Energy Physics (HEP), but it has since grown to include participants from diverse fields, such as medicine, astrophysics, and statistics.  Unsurprisingly, a workshop centered on this multidisciplinary approach to accelerated machine learning drew the participation of several members of A3D3.  And just as A3D3 revolves around mutual support and cross-disciplinary efforts, participants in the workshop were intrigued to see how the same techniques and algorithms were found in diverse applications across different disciplines.

“The workshop brings together researchers from very different specialties who do not typically have a chance to come together and exchange ideas,” said Fermilab’s Nhan Tran, one of the original FastML organizers.  “Despite this, it was so refreshing to see many amazing talks and enthusiastic discussion from all the workshop participants willing to get out of their comfort zones and expand their research.  I really appreciate that spirit and it makes the workshop series very unique and fun.”

Emphasizing the increased scope of the workshop series, Fermilab’s Kevin Pedro said that he attended the workshop “to learn about new cutting-edge computational techniques that are accelerating ML throughout many scientific fields.”

At the workshop, seven A3D3 trainees gave presentations on their work.  Leading off, Farouk Mokhtar of UCSD and Santosh Parajuli of UIUC presented their work on implementing machine learning models for the LHC experiments CMS and ATLAS, respectively.  Though working on independent efforts, both use graph neural networks to efficiently and scalably reconstruct particles.

Jeffrey Krupa presents his work on developing a Sparse Point Voxel CNNSantosh Parajuli presents his work on implementing graph neural networks for Event Filter Tracking in ATLAS.  In keeping with the “Fast” theme of the workshop, he is developing a VHDL implementation of his algorithm to run on FPGAs. for machine-learning-based clustering in hadronic calorimeters.

Next up, Patrick McCormack and Jeffrey Krupa, both of MIT, gave talks about two projects that they both work on.  One is a CMS effort to implement GPU acceleration for ML-methods via an Inference as a Service scheme, and the other is an implementation of a Sparse Point Voxel CNN (SPVCNN) for determining clusters of energy in hadron calorimeters for LHC experiments.  According to Krupa, “the SPVCNN algorithm is a first-time use of HCAL depth segmentation in clustering for CMS, and it removes the latency associated with HCAL clustering from reconstruction workflows.”

Jeffrey Krupa presents his work on developing a Sparse Point Voxel CNN for machine-learning-based clustering in hadronic calorimeters.

The last A3D3 talk from the workshop’s first day was given by Duc Hoang, also from MIT, who presented his work on algorithms for the CMS Layer-1 Trigger. These algorithms must be able to produce inferences at 40 MHz, such that one must balance algorithmic complexity with speed.  Thanks to the efforts of Duc and his collaborators, the algorithms that they have implemented on FPGAs for both bottom quark and tau lepton identification will increase the efficiency of the CMS trigger system for rare processes with these particles.

During the second day of the workshop, the focus shifted away from the LHC.  The gravitational waves side of A3D3 was represented by Katya Govorkova and Eric Moreno of MIT.  Katya presented her work on the development of Gravitational Wave Anomalous Knowledge (GWAK), a method for tagging gravitational waves from anomalous sources.  This algorithm is related to the Quasi-Anomalous Knowledge (QUAK) technique that was developed by A3D3 members for application in LHC contexts.  The GWAK algorithm has since evolved and can be used in real time to help distinguish between truly anomalous gravitational waves and meaningless detector glitches.

Katya Govorkova presents “Gravitation Wave Anomalous Knowledge”, or GWAK.  As she was giving the first talk about gravitational waves at the workshop, she covered the basic physics behind the LIGO experiment.

Eric’s talk focused on the Machine Learning for Gravitational Waves (ML4GW) package, which is a suite of tools enabling real-time machine learning for gravitational wave (GW) experiments, such as LIGO.  These tools have accelerated GW-detection, parameter estimation for events, noise regression, and anomaly detection via GWAK.

The workshop’s concluding remarks were given by A3D3’s deputy director and MIT professor Phil Harris.  He exhorted the audience with this playful paradox: “In order to go fast, we have to go slow.  By which I mean that designing an algorithm or workflow for the fastest possible performance takes time and careful consideration.”  In his 20 minute talk, he pointed out many of the similarities and common tools being used across disciplines, such as sparsification and quantization of neural networks, hardware-based acceleration using GPUs and FPGAs, and deep learning architectures such as transformers and graph neural networks.  A complete summary of the topics covered in the workshop would be far too long for this article, but the workshop’s timetable, along with links to most presentations can be found here.

“I really enjoyed that the workshop has a very diverse group of participants and talks that I found very inspiring,” said Professor Mia Liu from Purdue reflecting on the workshop.  “I am learning and thinking of new ways of accelerating science by developing appropriate algorithms and learning methods, in addition to my current research in real-time ML in low latency and high throughput systems.  For example, robust learning methods for embedding of scientific data, that can account for the variance due to the nature of the physical object and the measurement methods etc, is challenging but crucial for broader and long lasting impact of ML on scientific discoveries.”

The workshop also included tutorials on state-of-the-art deployment techniques, such as the hls4ml package for creating firmware implementations of machine learning algorithms, Intel AI Suite for deploying algorithms on Intel FPGAs, and the use of Intelligence Processing Units (IPUs) from Graphcore.  There were also informal tours of several of the labs at ICL.

On a lighter note, some workshop participants found time to explore the sights and sounds of London.  A3D3 members Eric and Duc also put together a public lecture from rapper Lupe Fiasco (Wasalu Jaco), who is currently a visiting professor at MIT.  He discussed some of his work with Google on creating TextFX, which is a large language model for exploring relationships between words and generating phonetically, syntactically, or semantically linked phrases.  They also managed to bring in Irving Finkel of the British Museum, who discussed the history of the game of Ur, which is a shared love with Lupe, Eric, and Duc.

Duc, Eric, Irving Finkel, and Lupe Fiasco discussed the game of Ur in a panel after Lupe’s public lecture on the relationship between rap and large language models.

The workshop proved to be a valuable experience for A3D3 attendees, and I suspect that future iterations will be well attended by our members.  “Attending the workshop was an incredible experience for me,” said Santosh Parajuli.  “The FastML workshop provided a unique platform to learn from experts, exchange ideas, and explore the latest advancements in different fields.  Additionally, I had a chance to share our exciting work on using advanced technology and machine learning to improve how we track particles in high-energy physics, which could help us make big discoveries in the future!”