首页|University of Notre Dame Researchers Add New Data to Research in Physics (Multi-resolution partial differential equations preserved learning framework for spatiotemporal dynamics)
University of Notre Dame Researchers Add New Data to Research in Physics (Multi-resolution partial differential equations preserved learning framework for spatiotemporal dynamics)
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
By a News Reporter-Staff News Editor at Network Daily News - Investigatorsdiscuss new findings in physics. According to news reporting originating from the University of Notre Dameby NewsRx correspondents, research stated, “Traditional data-driven deep learning models often strugglewith high training costs, error accumulation, and poor generalizability in complex physical processes.”Financial supporters for this research include United States Department of Defense | United StatesNavy | Office of Naval Research; National Science Foundation.Our news correspondents obtained a quote from the research from University of Notre Dame: “Physicsinformeddeep learning (PiDL) addresses these challenges by incorporating physical principles into themodel. Most PiDL approaches regularize training by embedding governing equations into the loss function,yet this depends heavily on extensive hyperparameter tuning to weigh each loss term. To this end,we propose to leverage physics prior knowledge by ‘baking’ the discretized governing equations into theneural network architecture via the connection between the partial differential equations (PDE) operatorsand network structures, resulting in a PDE-preserved neural network (PPNN). This method, embeddingdiscretized PDEs through convolutional residual networks in a multi-resolution setting, largely improvesthe generalizability and long-term prediction accuracy, outperforming conventional black-box models.”
University of Notre DameNetworksNeural NetworksPhysics
NETL
NSTL
