Half a Million Dollars FY 2019 DURIP Grant is Awarded to the Center for Informatics and Computational Science (CICS) for Artificial Intelligence (AI) Research applied to Scientific and Engineering Applications
The Air Force Office of Scientific Research (AFOSR) Computational Mathematics Program has awarded half a million dollars of the FY 2019 Defense University Research Instrumentation Program (DURIP) Grant to the Center for Informatics and Computational Science (CICS) for “Physics-Constrained Deep Learning research”. The funded proposal on “General Purpose GPU- Accelerated Computing Cluster for Data-Driven Multiscale-Multiphysics Uncertainty Quantification” would provide the resources for the acquisition of a GPU-based high-performance scientific computing cluster dedicated to Artificial Intelligence (AI) research conducted by CICS members. The annual DURIP award process is highly competitive receiving 669 proposals, requesting $259 million in funding.
The anticipated outcome of the proposed work is the development of data-driven methodologies rooted in Bayesian Deep Learning that allows for one to leverage “Big Data” in scientific applications to a better model, interpret, and propagate uncertainty in complex physical, chemical, and biological systems. Such efforts are essential to provide credibility to simulation-based predictive design and analysis and are ubiquitous in the aerospace, naval, energy, pharmaceutical, and life-sciences sectors. The proposed research will provide fundamental mathematical and statistical advances in Deep Learning methodologies for modeling and quantifying uncertainty in simulation and design of complex multiscale/multiphysics systems. Of particular interest are methodologies that embed physical constraints in the AI network architectures to guarantee generalization results that obey such constraints. The DURIP award provides CICS with unique capabilities to enhance its research on Physics of AI Research applied across the colleges of Engineering and the Sciences.
The GPU cluster will become available to the members of the CICS participating in Deep Learning research. These resources will enhance the GPU capabilities at the Notre Dame Center for Research Computing (CRC) allowing CICS members to pursue large scale AI research. In addition to strengthening the mission of CICS, these resources will also be essential in training Notre Dame students on the latest practices in the integration of Machine and Deep Learning research with Scientific Computing to address the most pressing and complex societal issues of our days.