By: Will Benoit, Cristina Andrade, Tyler Barna, Michael Davis, and Andrew Toivonen
April 24, 2025

The 2025 Transients from Space conference featured the novel efforts of A3D3 members working in the domain of multi-messenger astronomy (MMA). Below, members of Dr. Coughlin’s group at the University of Minnesota share their experiences attending this conference.

Pictured above are the attendees of the Transients from Space conference from the University of Minnesota. From left to right: Cristina Andrade, Michael Davis, Dr. Michael Coughlin, Will Benoit, Tyler Barna, and Andrew Toivonen

Will Benoit: The Transients from Space (TFS) workshop, which centered around time-domain astronomy and multi-messenger astrophysics, was well-organized and well-attended. Over the course of the three-day meeting, there were presentations from all corners of the astronomical transients community, as well as ample time for smaller group and individual discussions. Thanks to the flexibility of the workshop organizers, I had the opportunity to present the ml4gw library that our group has developed under the auspices of the MMA pillar of A3D3. This library contains the infrastructure we’ve developed to support our machine learning algorithms in gravitational-wave physics, and the hope was to communicate some of the lessons we’ve learned over the course of development. In a similar vein, I also presented on the broader technical and sociological hurdles we’ve faced in getting our algorithms to the stage of production deployment. It was informative to hear the thoughts and ideas of scientists from a domain with different challenges and requirements than my own, and I hope that the discussions we had will spur focused development on real-time machine learning tools for this area.

Will Benoit presents on “Deploying Machine Learning in Time-Domain Astronomy”

Cristina Andrade: Attending the Transients from Space conference at STScI (Space Telescope Science Institute) was a productive and energizing experience. My work spans gravitational waves, compact mergers and the development of transient detection metrics for the Vera C. Rubin Observatory. So, having a diverse array of transient scientists in one place made for incredibly productive conversations. The “Target of Opportunity” breakout session was a highlight, offering a rare chance to move the needle on how we coordinate large-scale, rapid-response observations across facilities in anticipation of upcoming missions. I also appreciated the thoughtful discussion in the machine learning session, especially around standardizing tools to streamline detection and around resource prioritization across missions. I had the chance to share updates and get targeted feedback from Rubin leadership on the LSST transient metrics I’m developing. I connected with teams from several international observatories affiliated with the GRANDMA collaboration (a telescope network), which I help to lead operational efforts for. The conference not only deepened existing relationships between my group and the wider transient field but also clarified technical and logistical challenges that weare actively working to address.

Tyler Barna: TFS was a very productive workshop; in just three days, it provided an overview of the current state of space-based observation and detailed the community’s plans for missions launching in the coming decade. Transients represent a unique challenge in astronomy – they generally occur without warning and fade from view on varying timescales. There is still much science to be done with early-time observations of essentially all classes of transients. Kilonovae, my field of study, occur at time scales even shorter than many other transients, on the order of days. As our estimates for their occurrence rates have developed, we have come to understand them to be relatively rare events, so any opportunity to observe them must be seized upon. Many of the missions discussed at TFS, such as the Nancy Grace Roman Space Telescope, have specifications that will be incredibly valuable for observing kilonovae. I hope the discussions at TFS surrounding collaboration in managing ToO (Target of Opportunity) and DDT (Director’s Discretionary Time) resources bear fruit so the community can maximize observation of rare events like kilonovae.

Michael Davis: The Transients from Space workshop at the Space Telescope Science Institute was a great opportunity to learn about the latest developments in time-domain and multi-messenger astronomy. Over three days, I was introduced to current and future space-based transient surveys, the challenges of rapid follow-up observations, and the role of upcoming missions like the Nancy Grace Roman Space Telescope. It was especially valuable to hear discussions on how to optimize Target of Opportunity strategies for rare and short-lived events. Beyond the talks, I enjoyed meeting others in the transient astronomy community and discussing shared challenges in data analysis and follow-up coordination.

Andrew Toivonen: As someone who works at the intersection of gravitational-wave searches and multi-messenger follow-up efforts, transient detection is a key aspect of my research. It was great to discuss with astronomers what is needed to make efficient and effective follow-up decisions in a rapidly evolving field. For example, while presenting my poster, I had an astronomer interested in GW (gravitational wave) follow-up approach me to comment on how such multi-messenger data products would be useful to their field. I also thought the machine learning discussions were positive. Driven by Will’s talk about ML4GW and a shared ML+GW library, there was discussion on whether a similar thing could be done for transient science. The main concern was the difficulty that many ML algorithms have with light curves, and the abundance and variety of training data that would be needed. This workshop has built momentum towards collaboration between different groups with common goals. 

By: Mark Neubauer
November 27, 2024

The 3rd annual conference of the NSF HDR Ecosystem was held on the campus of the University of Illinois Urbana-Champaign from the 9th to the 12th of September 2024

Attendees of the 2024 HDR Ecosystem Conference at the University of Illinois Urbana-Champaign

Venue

Conference Welcome and Overview

Dean Rashid Bashir of the Illinois Grainger College of Engineering kickoffs of conference with a welcome and vision
A3D3 Institute Director Shih-Chieh Hsu (U. Washington) describes the Institute’s Accomplishments, Activities, Plans

Keynote

Vipin Kumar (Minnesota) delivers the Keynote talk on Knowledge-Guided Machine Learning: A New Framework for Accelerating Scientific Discovery and Addressing Global Environmental Challenges

Evening Public Lecture

Suresh Venkatasubramanian (Brown University) delivers a Free and Open Public Lecture on AI Policy toward Making AI Safe, Effective and Trustworthy

Transdisciplinary and Cross-Cutting Research Breakouts

Participants joined discussion breakout sessions to discuss topics including LLMs / Foundation Models for Research, Responsible AI / Ethical AI, Knowledge-Guided ML / Physics-Informed Neural Networks, Continuous ML and Human-in-the-Loop Decision Making, Future ML challenges, Challenges and Opportunities for HDR Institutes and NAIRR Integration, and Interdisciplinary Careers

Lightning Talks and Poster Session

Participants give a 1-minute poster pitch talk on their topic before moving to the poster room to chat with others about their posters

Official Launch of the HDR Machine Learning Challenge!

A3D3 Deputy Director Phil Harris talks to the audience about the inaugural NSF HDR ML Challenge

Tour of the National Petascale Computing Facility

Brett Bode (National Center for Supercomputing Applications) gives conference attendees a tour of the computing facilities at Illinois. The DeltaAI supercomputer (center) was launched a month after the tour.

By: Kira Nolan
March 4, 2025

A new record for the largest gathering of astronomers was set this January, as around 3,700 people traveled to the 245th American Astronomical Society (AAS) meeting in National Harbor, Maryland. These biannual AAS meetings bring together scientists, engineers, educators, students and advocates from every corner of astronomy. I was able to attend this expansive conference for the first time, and will share my experience in this blog post.

1 – Everything, everywhere, all at once

Plenary talks ranged from how to crash a rocket into an asteroid (DART mission), to radio astronomy at the South Pole (South Pole Telescope), to planet accretion and best practices in research mentorship. At any given time, simultaneous sessions covered topics ranging from planetary science to cosmology. Even as a postbac, I tend to focus my attention on papers and work that I think will be directly helpful for my own projects. I intentionally chose to attend a mix of sessions that were either very relevant or completely disconnected from my work. Just like how A3D3 allows scientists to look outside of their domains, AAS is a great opportunity for young astronomers like myself to get a crash course on the bigger picture of work happening across the field. 

2 – Astronomy is big and small

Paradoxically for such a large conference, AAS makes the astronomy world feel small. My experience highlighted just how many connections the postbac has allowed me to make in the field, over a year of virtual collaboration with different groups and travel to conferences. I connected with people ranging from a graduate student I met the first week of my postbac to a professor I exchanged emails with regarding a research question. Browsing the conference exposition hall, I got to talk to the developer of software for the Fermi Gamma-Ray Space Telescope that I worked with. This exposure to the field is an invaluable part of the postbac experience.

3 – Presenting a poster

I presented a poster on my work automating the multi-messenger follow-up for binary black hole mergers. This was my first time presenting a poster outside of the A3D3 community, and the experience taught me lessons that I will be able to use for future poster sessions. As I talked with people ranging from undergraduate students to senior professors, from all different fields within astronomy, I got practice explaining my work.

4 – What about machine learning (ML)?

The AAS has recently established a task force focused on the rise of artificial intelligence (AI) in the field, and at this meeting, there were around ten different sessions explicitly dedicated to topics around ML in astronomy. These included discussions around developing astronomy datasets for machine learning challenges and using AI for advanced statistical inference. Some examples of work include efforts towards physics-informed AI for astronomy and high-dimensional inference for astronomical image reconstruction. While ML has long been applied to astronomy datasets, astronomers are faced with growing data streams and are interested in accessing the cutting edge of machine learning to most efficiently utilize those data for exciting discoveries.

By: Katrine Kompanets
March 4, 2025

The Conference for Undergraduate Women and Gender Minorities in Physics (CU*iP) is a unique annual event that takes place at multiple locations across the country simultaneously. This January, A3D3 was represented by members attending multiple locations of the conference.

Pictured from left to right: Megan Averill, Katrine Kompanets, Emma de Bruin, and Yiwen Chen attend CU*iP at Michigan Tech.

Students from the University of Minnesota attended CU*iP at Michigan Tech, where they got to explore the high-end laboratories there and network with fellow undergraduates. Yiwen Chen and Megan Averill presented “Distinguishing Astrophysical Signals from Noise: Machine Learning for Gravitational Waves Detection,” Emma DeBruin presented “Improving Sensitivity to Neutron Star Gravitational Wave Events using the Qp Transform,” and Katrine Kompanets shared her research on “Improving Sensitivity of Gravitational Wave Event Detections Using Machine Learning”. It was an amazing experience to meet students from all over the country and share experiences and research with each other!

At the University of California, Berkeley, students at the CU*iP career fair enjoyed learning about the A3D3 Postbaccalaureate Research Fellowship and the many different research areas and synergies under the A3D3 umbrella. Quite a few students expressed enthusiasm for applying to the postbac program, and a few even said it was already their top choice after graduation!

A3D3 members operate a booth at the UC Berkeley CU*iP career fair.

By: Eli Chien
December 19, 2024

Vancouver, Canada – The Thirty-Eighth Annual Conference on Neural Information Processing Systems (NeurIPS 2024) was held in Vancouver from December 10 to 15, 2024. NeurIPS is one of the premier machine learning and AI conferences, gathering outstanding researchers from around the world. A3D3 members contributed seven posters, including one spotlight poster and one benchmark poster, and three workshop papers. In addition, A3D3 member Dr. Thea Klaeboe Aarrestad from ETH Zürich was an invited speaker at the WiML (Women in Machine Learning) workshop and the ML4PS (Machine Learning and the Physical Sciences) workshop, presenting “Pushing the limits of real-time ML: Nanosecond inference for Physics Discovery at the Large Hadron Collider.” We are excited to see that A3D3 members are an important part of the machine learning community!

A3D3 members gather at NeurIPS 2024.

Invited Talks

Dr. Thea Klaeboe Aarrestad gives invited talks for the WiML workshop (left) and the ML4PS workshop (right).

Poster Presentations

Mingfei Chen (left) and Dr. Eli Shlizerman (right) from the University of Washington presented “AV-Cloud: Spatial Audio Rendering Through Audio-Visual Cloud Splatting.” They propose a novel approach, AV-Cloud, for rendering high-quality spatial audio in 3D scenes that is in synchrony with the visual stream but does not rely on nor is explicitly conditioned on the visual rendering. [paper link, project link]

Dr. Eli Chien from the Georgia Institute of Technology presented a spotlight poster titled “Langevin Unlearning: A New Perspective of Noisy Gradient Descent for Machine Unlearning.” They propose Langevin unlearning which not only connects DP with unlearning, but also provides unlearning privacy guarantees for PNGD with Langevin dynamic analysis. [paper link]

Eli Chien also presented a poster titled “Certified Machine Unlearning via Noisy Stochastic Gradient Descent.” They propose to leverage stochastic noisy gradient descent for unlearning and establish its first approximate unlearning guarantee under the convexity assumption. [paper link]

Jingru (Jessica) Jia from the University of Illinois at Urbana-Champaign presented “Decision-Making Behavior Evaluation Framework for LLMs under Uncertain Context.” This paper quantitatively assesses the decision-making behaviors of large language models (LLMs), demonstrating how these models mirror human risk behaviors and exhibit significant variations when embedded with socio-demographic characteristics. [paper link]

Shikun Liu from the Georgia Institute of Technology presented “GeSS: Benchmarking Geometric Deep Learning under Scientific Applications with Distribution Shifts.” They propose GeSS, a comprehensive benchmark designed for evaluating the performance of geometric deep learning (GDL) models in scientific scenarios with various distribution shifts, spanning scientific domains from particle physics and materials science to biochemistry. [paper link]

Yingbing Huang from the University of Illinois Urbana-Champaign presented “SnapKV: LLM Knows What You are Looking for Before Generation.” They propose SnapKV, a fine-tuning-free method to reduce the Key-Value (KV) cache size in Large Language Models (LLMs) by clustering important positions for each attention head. [paper link]

Xiulong Liu from the University of Washington presented “Tell What You Hear From What You See – Video to Audio Generation Through Text.” It is a novel multi-modal generation framework for text guided video-to-audio generation and video-to-audio captioning. [paper link]

Workshop Presentations

Zihan Zhao from the University of California, San Diego presented “Learning Symmetry-Independent Jet Representations via Jet-Based Joint Embedding Predictive Architecture” at ML4PS workshop. They propose an approach to learning augmentation-independent jet representations using a jet-based joint embedding predictive architecture (J-JEPA). [paper link, project link]

Trung Le from the University of Washington presented “NetFormer: An interpretable model for recovering identity and structure in neural population dynamics” at the NeuroAI workshop. They introduce NetFormer, an interpretable dynamical model to capture complicated neuronal population dynamics and recover nonstationary structures. [paper link]

Andrew “AJ” Wildridge from Purdue University presented “Bumblebee: Foundation Model for Particle Physics Discovery” at ML4PS workshop. Bumblebee is a transformer-based model for particle physics that improves top quark reconstruction by 10-20% and effectively generalizes to downstream tasks of classifying undiscovered particles with excellent performance. [paper link, project link]

Patrick Odagiu from ETH Zurich presented “Knowledge Distillation for Teaching Symmetry Invariances” at SciforDL Workshop. They show that knowledge distillation is just as good as data augmentation for learning a specific symmetry invariance in your data set. [paper link]

By: Cymberly Tsai
December 19, 2024

About the author: Cymberly Tsai is a high school senior at Council Rock High School North working with Prof. Joshua Agar at Drexel University. Cymberly presented a poster titled “Foundation Model for Real-Time Model Selection and Fitting” at the A3D3 All-Hands Meeting at Purdue University.

Participants of the 2024 A3D3 All-Hands Meeting gather for a group photo.

The A3D3 Institute hosted its annual All-Hands Meeting at Purdue University on October 14, 2024. The conference was a constellation of ideas as participants presented bright projects, which contributed to a sky of innovation, and guided others at various stages of education. As a student at Council Rock High School North, I had the opportunity to attend this conference with the Drexel Multifunctional Materials and Machine Learning lab with professor Josh Agar, after working with the lab the past summer. Taking place at the intersection of artificial intelligence and scientific research, the conference included attendees from a vast range of universities and locations, offering both in-person and virtual participation.

The conference emphasized accelerating machine learning for applications in a variety of fields. Located at Purdue’s Convergence Center, the meeting began with talks from several principal investigators, including Dr. Song Han‘s talk introducing efficient multi-modal large language model (LLM) innovations and Dr. Pan Li’s work on Graph Machine Learning under distribution shifts. These in-depth presentations offered insight into specific innovations, inspiring further research from attendees. Highlighting many varied applications of machine learning, trainees presented overviews of fields with high potential for fast machine learning, such as multi-messenger astronomy, neuroscience, high-energy physics, and hardware/algorithm co-development. The trainee-led talks covered such cutting-edge topics as the challenges of processing massive data rates in high-energy physics research and real-time brain modeling in primates with machine learning accelerated by Field-Programmable Gate Arrays (FPGAs). As one of the five National Science Foundation’s (NSF) Harnessing the Data Revolution (HDR) institutes, A3D3 also announced the NSF HDR ML Anomaly Detection Challenge at the conference, which leverages datasets in multiple fields to encourage data-driven discovery.

The All-Hands Meeting was particularly beneficial for those early in their research careers, welcoming them into the academic community. The conference offered a poster session, providing a chance to engage with ideas in a low-stress setting. Beyond research-specific content, the conference included a panel discussion on how to apply to faculty, postdoc, and industry jobs, led by Dr. Zhijian Liu, an NVIDIA research scientist and incoming assistant professor at UCSD, Dr. Julia Gonski, a Panofsky Fellow at SLAC/Stanford, and Dr. Dylan Rankin, an assistant professor at the University of Pennsylvania. Postbaccalaureate fellows of A3D3 also described their experience, giving insight into another avenue of research. A highlight of the All-Hands Meeting was the trainee-led townhall, which gave trainees and early-career researchers a chance to discuss ideas such as potential presentation opportunities and future collaborations.

The A3D3 conference experience expanded my understanding of machine learning and the world of research, sparking a greater curiosity and motivation to pursue scientific advancements with artificial intelligence. Following the All-Hands Meeting, the Fast Machine Learning for Science Conference conveniently took place at Purdue from October 15 through 18, allowing A3D3 attendees to dive deeper into machine learning advancements in research. Attending the conference was an invaluable experience for me to engage in the world of research and witness innovations in machine learning that I hope to contribute to in the future.

By: Miles Cochran-Branson
July 26, 2024

Following the three-day US-ATLAS conference held at the University of Washington (UW) Seattle campus, students met for a brief tutorial on tools available for US-ATLAS members. After introductions in computing resources, Yuan-Tang Chou—a postdoc at UW and member of the A3D3 team—gave a presentation on GPU resources and utilization with a focus on new applications to particle-physics workflows. 

Chou gives a brief presentation on applications of the NVIDIA Triton server for deploying models as-a-service.

The presentation focused on deploying models on accelerators such as GPUs, as-a-service (aaS) using the NVIDIA Triton Inference Server. Chou discussed the merits of heterogeneous computing—the most straightforward way to deploy algorithms where the CPU and GPU both are connected on a single node. He noted that for many physics processes such as Graph-Neural-Network (GNN) based tracking, flavor tagging, and detector simulation, heterogeneous computing could be “inefficient and very expensive to scale.” Hence, offloading expensive tasks to a GPU server could streamline the deployment of large models important for ATLAS physics. 

Sample architecture of as-a-service model deployment from CPU-only client nodes or CPU / GPU client nodes to a GPU server.

After motivating why deploying models as-a-service could be beneficial in physics analysis, Chou gave a brief demo on deploying GNN-tracking aaS. This was followed by a hands-on tutorial deploying the resnet50 image recognition deep neural network as-a-service on computing resources at CERN. The tutorial material focused on building the proper model repository structure and configuration for image detection on a GPU server. Students set-up a work environment, deployed a backend on the server, and sent an image to the server to be classified. 

Students work on deploying a backend on a server and sending images to the server to be classified. 

By the end of the tutorial, students had successfully deployed a backend and received image classifications back from the image recognition model. Interested students were connected with experts currently working in ongoing development of aaS tools in algorithm development of ATLAS algorithms.

Tutorial Resources: https://hrzhao76.github.io/AthenaTriton/triton-lxplusGPU.html. Developed by Yuan-Tang Chou, Miles Cochran-Branson (UW), Xiangyang Ju (LBNL), and Haoran Zhao (UW)

Written by Miles Cochran-Branson, PhD student at University of Washington