NSF-Simons CosmicAI Institute Seminar Series with Varun Shankar - March 26

Event Information

March 26, 2025

12:00 PM – 1:00 PM

Evans Conference Room, Warnock Engineering Building (WEB) Room 3780

Zoom Access

Meeting ID: 958 8067 9001

Passcode: 255836

Structure-Preserving, Low-Parameter, Interpretable, Operator Learning for Surrogate Modeling with Varun Shankar (Assistant Professor, Kahlert School of Computing)

Scientific machine learning (SciML) is a relatively new scientific discipline that weds scientific computing and high performance computing with carefully designed machine learning (ML) techniques. In the context of astrophysics, SciML has been applied to galaxy classification and identification, outlier detection, and uncertainty quantification.

Within SciML, operator learning is a rapidly emerging and powerful new paradigm for surrogate modeling across engineering and the sciences, with recent successes in climate modeling, material design, and carbon sequestration problems (to name a few). In this talk, I will present a unified framework that encompasses many operator learning paradigms and use this to present three advancements in operator learning: (1) the Kernel Neural Operator (KNO), a generalization of the Fourier neural operator that allows for greater flexibility in kernel choices and for local spatial adaptivity while inherently using far fewer trainable parameters; (2) the ensemble DeepONet, a generalization to Deep Operator Networks that enables the incorporation of spatial adaptivity directly into a set of basis functions; and (3) a new operator learning paradigm based on kernel approximation that analytically preserves the divergence free property and requires minimal training, all while achieving state-of-the-art performance on incompressible flow problems.

We argue that operator learning has the potential to positively impact astrophysics through trustworthy, rapid, and interpretable surrogate models for multiscale simulations of magnetohydrodynamics (MHD) and numerical general relativity (GR), and for inverse problems such as physical parameter estimation.

by Ellen Lewis

UCBPC Computing Industry Panel—February 11

Tomorrow (February 11) at 3:30 PM, UCBPC hosts their Computing Industry Panel in Marriott Library Room 1120. Join us for an engaging discussion with professionals from top tech companies and academia. We hope to see you there!

by Ellen Lewis

Guest Speaker: Proximal Causal Inference with Text Data talk by Katie Keith

When: Monday, January 27, 2–3:30pm

Where: WEB 3780 (Evans Conference Room)

Talk Title: Proximal Causal Inference with Text Data

Speaker’s Bio: Katherine (Katie) Keith is currently an Assistant Professor in the Computer Science department at Williams College in Massachusetts. Her research interests are at the intersection of natural language processing, causal inference, and  computational social science. Previously, she was a Postdoctoral Young Investigator at the Allen Institute for Artificial Intelligence, and she graduated with a PhD from the Manning College of Information and Computer Sciences at the University of Massachusetts Amherst. She has been a co-organizer of the First Workshop on Causal Inference and NLP, co-organizer of the NLP+CSS Workshops at EMNLP at NAACL, and was a recipient of a Bloomberg Data Science PhD fellowship.

Talk Abstract:

Causal inference underlies many important policy decisions and interventions. For example, clinicians must decide how to prescribe medications to patients, central bank committees must decide how to change interest rates, and online platform administrators must decide how to moderate users. In the absence of a randomized controlled trial, one can turn to observational (non-experimental) data to estimate causal effects. In this setting, a primary obstacle to unbiased causal effect estimation is confounding variables, variables that affect both the treatment (e.g., which medication) and the outcome (e.g., an aspect of patient health). In many applications, a rich, unstructured source of confounding variables is text data: notes from electronic health records (EHRs) detail patients’ personal and medical histories, newspaper articles document national and international events, and online platforms host exchanges of users’ written opinions. By expanding observational causal estimation methods that can incorporate natural language processing (NLP) and text data, analysts may be able to make inferences in a wider range of settings. 

Recent text-based causal methods attempt to mitigate confounding bias by estimating proxies of confounding variables that are partially or imperfectly measured from unstructured text data. These approaches, however, assume analysts have supervised labels of the confounders given text for a subset of instances, a constraint that is sometimes infeasible due to data privacy or annotation costs. In this work, we address settings in which an important confounding variable is completely unobserved. We propose a new causal inference method that uses two instances of pre-treatment text data, infers two proxies using two zero-shot models on the separate instances, and applies these proxies in the proximal g-formula. We prove, under certain assumptions about the instances of text and accuracy of the zero-shot predictions, that our method of inferring text-based proxies satisfies identification conditions of the proximal g-formula while other seemingly reasonable proposals do not. To address untestable assumptions associated with our method and the proximal g-formula, we further propose an odds ratio falsification heuristic that flags when to proceed with downstream effect estimation using the inferred proxies. We evaluate our method in synthetic and semi-synthetic settings—the latter with real-world clinical notes from MIMIC-III and open large language models for zero-shot prediction—and find that our method produces estimates with low bias. We believe that this text-based design of proxies allows for the use of proximal causal inference in a wider range of scenarios, particularly those for which obtaining suitable proxies from structured data is difficult.

by Ellen Lewis