The Artificial Intelligence Graduate Group (AIGG) welcomes Derek Ferguson, Chief Software Officer of the Fitch Group, for an AI industry talk.

Data Augmentation for Graph Regression

Meng Jiang, Associate Professor, Department of Computer Science and Engineering, University of Notre Dame

Abstract: Graph regression plays a key role in materials discovery by enabling the prediction of numerical properties of molecules and polymers. However, graph regression models often rely on training sets with only a few hundred labeled examples, and these labels are typically imbalanced. While a large number of unlabeled examples are available, they are often drawn from diverse domains, making them less effective for improving target label predictions. In machine learning, data augmentation refers to techniques that increase the size of the training set by generating slightly modified or synthetic versions of existing data. These methods are simple yet effective. In this talk, I will introduce three graph data augmentation techniques tailored for supervised learning, imbalanced learning, and transfer learning in graph regression tasks. The first technique leverages the mutual enhancement between model rationalization and data augmentation, improving both accuracy and interpretability in molecular and polymer property prediction. This approach demonstrates that graph data augmentation can be effectively performed in latent spaces. The second technique generates representations of additional data points with underrepresented labels to balance the training set. The third technique introduces a graph diffusion transformer (Graph DiT) that facilitates data-centric transfer learning, addressing the limitations of self-supervised methods when dealing with unlabeled graph data. Graph DiT integrates multiple properties such as synthetic score and gas permeability as condition constraints into diffusion models for multi-conditional polymer generation. Lastly we will discuss foundation model approaches for materials discovery.

Related Info

Title: Two Tales, One Resolution: Physics-Informed Test Time Scaling and Precondition

Speaker: Yiping Lu, Northwestern University

Abstract: In this talk, I will introduce a novel framework for physics-informed debiasing of machine learning estimators, which we call Simulation-Calibrated Scientific Machine Learning (SCaSML). This approach leverages the structure of physical models to achieve two key objectives:

Unbiased Predictions: It produces unbiased predictions even when the underlying machine learning predictor is biased.
Overcoming Dimensionality Challenges: It mitigates the curse of dimensionality that often affects high-dimensional estimators.

The SCaSML paradigm integrates a (potentially) biased machine learning algorithm with a de-biasing procedure that is rigorously designed using numerical analysis and stochastic simulation. Our methodology aligns with recent advances in inference-time computation—similar to those seen in the large language model literature—demonstrating that additional computation can enhance ML estimates. Furthermore, we establish a surprising equivalence between our framework and another research direction that utilizes approximate (linearized) solvers to precondition iterative methods. This connection not only bridges two distinct areas of study but also offers new insights into improving estimation accuracy in complex, high-dimensional (PDE) settings.

Zoom: https://northwestern.zoom.us/j/94570889326

-----

To subscribe to the Applied Mathematics Colloquia List send a message to LISTSERV@LISTSERV.IT.NORTHWESTERN.EDU with the command:

SUBSCRIBE esam-seminar FirstName LastName

Members of the NITMB community are invited to join us for Research-In-Progress meetings, an informal venue for members of the NITMB to discuss ongoing and/or planned research.

Xueying Wang is a Postdoctoral Fellow at the National Institute for Theory and Mathematics in Biology. Earning her Ph.D. in Physics from the University of Illinois, Urbana-Champaign, her research tackles the dynamical properties of complex, chaotic, and out-of-equilibrium systems, including fluid turbulence, biological and artificial neural networks, ecological systems, and active matter. In her doctoral work, she developed a spatially extended stochastic ecological model of energy flow in a fluid undergoing the transition to turbulence and predicted the four different phases encountered during the progression to fully developed turbulence in the quasi-one-dimensional flow. She employs a combination of computational and analytical techniques derived from statistical physics in her research. She has widespread research interests ranging from fluid turbulence to generalized learning & adaptation and structural stability & emergent functionality. For more information, see personal website and Google Scholar page.

Learn more about Xueying Wang's research and engage in discussion with the NITMB community. Research-In-Progress talks take place on Wednesdays at 3pm at the NITMB office (875 N Michigan Ave., Suite 4010). Snacks and coffee will follow.

Related Info

The Department of Sociology's Urban & Community Workshop presents:

Evie Lopoo (NU): Understanding Creative Labor in the Indie MusicScene: Competition and Community in the Age of AI

Sophia Costa (U Chicago): Responding to Poverty Ungovernance During the COVID-19 Pandemic

(A. Fahlberg, C. Martins, A. C. Araujo, L. Santos & J. L. da Silva)

Parkes Hall, Room 222 | Zoom

4:00 PM to 5:30 PM

Related Info

2024-2025 Commencement Ceremony

Related Info

McCormick School of Engineering PhD Hooding and Master's Degree Recognition Ceremony. The most up to date information can be found on our graduation webpage.

McCormick School of Engineering Undergraduate Convocation. The most up to date information can be found on our graduation webpage.

Related Info

The Open SkAI 2025 conference is the U.S. National Science Foundation and Simons Foundation-funded SkAI Institute's inaugural conference. The main aim of the conference is to enhance and generate new Astro-AI research directions. Conference themes will include: Astro-AI research across survey astronomy, from stars and transients to galaxy formation, evolution, and the dark sector. We encourage researchers from all levels to apply to attend.

Related Info

The Open SkAI 2025 conference is the U.S. National Science Foundation and Simons Foundation-funded SkAI Institute's inaugural conference. The main aim of the conference is to enhance and generate new Astro-AI research directions. Conference themes will include: Astro-AI research across survey astronomy, from stars and transients to galaxy formation, evolution, and the dark sector. We encourage researchers from all levels to apply to attend.

Related Info