arXiv stat.ML
· Papers
Low-Rank Evolutionary Deep Neural Networks via Adaptive Tangent-Space Reduction
arXiv:2509.16395v2 Announce Type: replace Abstract: Evolutionary deep neural networks (EDNNs) solve time-dependent partial differential equations by evolving the neural-network parameters sequentially in time through a local least-squares problem. Their main computational bottleneck is that each time step requires the