Name: Natalie Isenberg
Pronouns: She/her/hers
Biography:
Natalie is an Amalie Emmy Noether Postdoctoral Fellow in the Applied Mathematics group at Brookhaven National Laboratory’s Computational Science Initiative (CSI). She obtained her Ph.D. in Chemical Engineering from Carnegie Mellon University in 2021. Natalie’s research interests are generally in applying and developing mathematical optimization methods for complex decision-making problems under uncertainty.
Institution/Lab: Brookhaven National Laboratory
Website: natalieisenberg.com
SRP Collaboration Topic/Title: Modeling and Control of Particle Accelerator Beams using Bayesian Neural Networks
Field or research area: Surrogate modeling/AI/ML
Please select all the topical areas that apply to your project:
Machine Learning and AI
Brief Abstract:
Particle accelerators are a central tool for scientific discovery, from unraveling the governing laws of fundamental particles, to understanding the universe’s origins. Brookhaven National Laboratory is home to the Relativistic Heavy Ion Collider (RHIC) and the future Electron Ion Collider (EIC), two world-leading particle accelerators. At RHIC, which is the largest particle accelerator in the United States, scientists are currently studying the properties of subatomic particles. Specifically, researchers use particle collisions to recreate extreme conditions like those shortly after the Big Bang: the prevailing theory for the origins of our universe. Both cutting-edge accelerator experiments rely on complex particle beam controls (e.g., altering trajectories) to create collisions that generate useful data. Ensuring the accuracy and reliability of particle collision experiments is of critical importance and requires the development of robust beam control methodologies. In this research project, we will build a data-driven model of beam position using Bayesian neural networks (BNNs). BNNs are machine learning models that provide accurate predictions while also quantifying the error inherent in the data, ensuring robust and reliable beam position forecasts. The student intern working on this project will gain practical experience in cutting-edge machine learning topics as applied to a real high-energy physics application.
Desired relevant skills, background, or interests:
Some experience in Python or Julia preferred, but not required.
Other comments:
Do any special requirements apply? In-Person Only; Permanent Resident OK; International OK
Other, specify:
Keywords:
particle accelerators; machine learning; bayesian neural networks; optimal control; uncertainty quantification
Lightning Talk Title: Building Confidence in the Unknown: Optimal Decision Making under Uncertainty