Research

Scientific Machine Learning for Universal EOS Modeling and Non-Interacting Free-Energy Functional Prediction

DoE/University of Rochester Award (~$50K, October 2025- September 2026)

PI: Xumin Liu

Develop a conditional neural process (CNP) model to predict the non-interacting free-energy functional. Since collecting the ground-truth KS-DFT values for higher temperatures is much more expansive, we will leverage few-shot learning and uncertainty quantification techniques and integrate them with relevant physics properties in novel ways to combat limited training data for accurate predictions of the difference between KS-DFT and OF-DFT calculations.

Overcoming Programming Barriers for Non-Computing Majors in Data Science

NSF IUSE Award (~$750K, June 2024- May 2027).

PI: Xumin Liu, Co-PIs: Erik Golen, Victor Perotti, Deepak Kumar (Bryne Mawr College), Chunmei Liu (Howard University), Senior Personnel: Feng Cui, Stephanie Godleski

The overarching goal of this project is to provide effective curricular materials to overcome the programming barriers, expose students to various data science topics, and teach them how to solve data problems in the context of their own disciplines. The project will: (1) develop an integrated learning platform to support both teaching and learning; (2) develop a set of course modules covering important data science topics with hands-on assignments designed for different disciplines; (3) deploy and evaluate the platform and course modules at three participating institutions including an HBCU and a women's liberal arts college; and (4) conduct a study to investigate if the impact of the developed platform and course modules on student learning is independent from students' prior computing experience, discipline, gender, and demographics.

A Multimodal Dynamic Bayesian Learning Framework for Complex Decision-making

DoD/ONR Award (~$1.6M, October 2018- September 2023)
PI: Qi Yu; Co-PIs: Daniel Krutz and Yu Kong

Machine Learning Data Model

Develop advanced machine learning models to support decision-making in military operations while addressing the following challenges:

Coordinating multiple concerns and missions (e.g., self-protection, position assets, resource allocation…)
Missions carry some form of risk and uncertainty
Data come in heterogeneous formats (sensor readings, transcribed communications, image/videos from surveillance) in real-time, and fast-changing
The decision model should provide interpretable recommendation and effectively incorporate human feedback

Utilizing Synergy between Human and Computer Information Processing for Complex Visual Information Organization and Use

NSF IIS Award (~$500K, July 2018- June 2023).
PI: Qi Yu; Co-PIs: Anne Haake, Rui Li, and Pengcheng Shi

The primary focus of this project is on complex image understanding in specialized domains. (e.g., medicine, bioinformatics, and so on) This proposed research aims to integrate human and computer capabilities to discover image semantics by (1) encoding image inspection and analysis behaviors that represent domain expertise, and (2) algorithmically fusing human expertise with image data.

Trans-disciplinary approaches for knowledge elicitation and extraction

Eye movement sequence and verbal narrative

Multimodal knowledge fusion

Using Novel Scientific Machine Learning to Revolutionize Computational Methods for High-Energy-Density Physics

DoE/University of Rochester Award (~$100K, October 2021- September 2023)
PI: Qi Yu

Develop a novel scientific machine learning framework to infer a universal non-interacting free-energy density functional from large-scale Kohn-Sham Density-Functional Theory (KS-OFT) data. It will enable orbital-free OFT (OF-OFT) for accurate and efficient understanding of dynamic material properties and provide essential scientific tools for discovery of function materials at extreme conditions.