Project Grant 2543755

The University of Chicago received a $342,196 Project Grant from the National Science Foundation's Division of Computing and Communication Foundations under the Computer and Information Science and Engineering program (CFDA 47.070). The grant, awarded on August 1, 2026, and scheduled for completion by July 31, 2031, supports research in adaptive experimental design and active representation learning. The project delivers a novel unified learning framework that integrates data representation and experiment selection, enabling artificial intelligence systems to not only model complex data but also autonomously determine which experiments or measurements to conduct. This addresses a critical challenge in scientific discovery and engineering design where experimental resources are limited and costly.

The research outputs include learning-based acquisition strategies that leverage representation learning, probabilistic modeling, and sequential decision-making techniques to optimize experiment selection in high-dimensional settings. The project develops uncertainty-aware modeling architectures, algorithms that incorporate multi-fidelity data sources and parallel experimentation, and decision policies trained using simulation and historical data. These technical innovations are designed to accelerate scientific discovery, improve engineering system efficiency, and enable intelligent decision-making in resource-constrained environments. Applications include optimization of scientific simulations and cyber-physical systems, demonstrating broad utility across multiple research domains where data collection costs or constraints significantly impact research feasibility.