Project Grant 2513966

This Project Grant award from the National Science Foundation's (NSF) Mathematical and Physical Sciences program (CFDA 47.049) provides $307,266 to North Carolina State University (NC State) to develop effective computational methods for training neural networks. The project aims to address fundamental challenges in artificial intelligence by creating a novel Exploration-Exploitation-Determination (EED) framework to significantly improve the training performance of neural networks, which are core components of modern AI models. The research objectives include establishing the EED framework for two-layer neural networks, developing a layer-wise training strategy for deep neural networks, extending the framework to handle noisy and corrupted data, and validating the methods through scientific machine learning applications. The project will make the resulting computational algorithms publicly available to enhance reproducibility and impact across scientific disciplines. It also includes educational and public engagement components, such as a summer research program for undergraduates and professional training for K-12 educators on computational mathematics for AI.