Sparse POD Mode Selection and Manifold Dimensionality Reduction with Neural Networks
稀疏POD模态选择与神经网络流形降维
机构 * School of Aerospace Engineering, Georgia Institute of Technology(航空航天工程学院,佐治亚理工学院) ; Computational Science Center, National Laboratory of the Rockies(岩石高原国家实验室计算科学中心)
AI总结 提出SparseModesNet框架,通过LassoNet实现线性POD模态的稀疏选择与非线性神经网络解码,在平流主导和混沌流中降低重构误差51-78%。