On April 18, the Center for Statistics and Machine Learning and the Princeton Institute for Computational Science and Engineering hosted their annual joint research colloquium. Throughout the afternoon, eight students from the certificate programs in Statistics and Machine Learning or Computational Science and Engineering presented their research. 

The joint colloquium is an opportunity to showcase computational research methods, statistical and machine learning analysis, and their applications across various research areas. “Computational methods, statistics, and machine learning are making an impact on every discipline,” said Michael Mueller, associate chair and professor of mechanical and aerospace engineering and director of the graduate certificate in computational science and engineering.

“The joint colloquium showcases the diversity of research at the university that intersects with computing,” said Mueller. “Each of the eight speakers came from a different department across engineering, natural sciences, and social sciences.”

Diversity of Research

This year’s certificate students came from various departments across campus, including Operations Research and Financial Engineering (ORFE), Atmospheric and Oceanic Sciences, the School of Public and International Affairs (SPIA), and Physics. Each participant was given 20 minutes to present their research.

From the department of Mechanical and Aerospace Engineering, Dario Panici presented on using a 3D equilibrium code for optimizing stellarator design. Nuclear fusion requires extremely high temperatures. To contain such hot plasma, researchers must employ magnetic confinement. In his work, Panici used the DESC code to explore optimal stellarator design for this magnetic confinement. 

In her presentation, Emma Levin, from the Department of Atmospheric and Oceanic Sciences, argued that machine learning and AI-based weather forecast systems are “the frontier of hurricane prediction.” Her goal is to use model-based research to uncover insights about how hurricane patterns will change in an increasingly warming world. 

Peter Kirgis from SPIA investigated whether various large language models exhibited political leanings. He devised a series of questions to evaluate a responder’s moral foundations and fed them to 16 different state-of-the-art LLMs. Kirgis then analyzed the models’ answers to find whether they leaned toward liberal or conservative values. 

Presenters Jack Draney, Isabel Moreira de Oliveira, Aardhya Pandey, Kehan Cai, and Jiayi Hu also elaborated on how computational and machine learning methods have advanced their understanding of research questions in their prospective fields. 

“One goal of the colloquium is to foster exchange across departments,” said Mueller. “This year’s colloquium provided a great example with two students in different departments working with molecular dynamics, exchanging best practices and new ideas.”

“This event helps build a community of scholars across related certificate programs, allowing students to learn from each other’s approaches and results,” said Peter Ramadge, Director of the Center for Statistics and Machine Learning. “This adds important value to their experiences as Princeton students.”