Project Grant 2212442

The National Science Foundation Office of Advanced Cyberinfrastructure awarded the University of Texas at Austin a $1.2 million Project Grant under the Computer and Information Science and Engineering program (CFDA 47.070) from September 1, 2022 to August 31, 2025.

The grant funds research to develop a rigorous and reliable scientific deep learning framework for forward, inverse, and uncertainty quantification problems in computational science and engineering. Specific objectives include equipping deep learning with underlying mathematical models to improve generalization, achieving comparable accuracy to traditional computational methods at lower cost, and quantifying uncertainty in deep learning solutions. The university will pursue these objectives through adaptive architecture design methods, model-constrained approaches to encode mathematical models into neural networks, and statistical techniques for quantifying neural network prediction uncertainty. Practical seismic wave propagation will serve as a testbed for developments. Findings will be incorporated into graduate coursework and open-sourced on TensorFlow, JAX and FEniCS/Firedrake to benefit the research community.