SPAMoE reduces average MAE by 44.4% on ten OpenFWI sub-datasets via a spectral-preserving DINO encoder plus frequency-routed MoE of FNO, MNO and LNO experts.
Ensemble and Mixture-of-Experts DeepONets For Operator Learning
4 Pith papers cite this work. Polarity classification is still indexing.
4
Pith papers citing it
representative citing papers
AMORE develops an adaptive multi-output DeepONet with custom losses, partition-of-unity trunk, and invertible/softmax mass-fraction maps to surrogate stiff kinetics on syngas (12 states) and GRI-Mech (24 states).
KL-DNN uses low-rank SVD and nested Karhunen-Loeve expansions to enable scalable operator learning on large 3D GCS simulations, achieving 0.04% relative pressure error and two-order speedup over DeepONet.
A neural operator framework embeds a snapshot-derived multiscale basis as trunk to learn permeability-to-solution mappings for nonlinear flow, avoiding online coarse nonlinear solves.
citing papers explorer
-
SPAMoE: Spectrum-Aware Hybrid Operator Framework for Full-Waveform Inversion
SPAMoE reduces average MAE by 44.4% on ten OpenFWI sub-datasets via a spectral-preserving DINO encoder plus frequency-routed MoE of FNO, MNO and LNO experts.
-
AMORE: Adaptive Multi-Output Operator Network for Stiff Chemical Kinetics
AMORE develops an adaptive multi-output DeepONet with custom losses, partition-of-unity trunk, and invertible/softmax mass-fraction maps to surrogate stiff kinetics on syngas (12 states) and GRI-Mech (24 states).
-
A Trainable-by-Parts Operator Learning Framework: Bridging DeepONet and Karhunen-Loeve Expansions for Large-Scale Applications
KL-DNN uses low-rank SVD and nested Karhunen-Loeve expansions to enable scalable operator learning on large 3D GCS simulations, achieving 0.04% relative pressure error and two-order speedup over DeepONet.
-
Operator Learning on the Data-Driven Multiscale Space for Nonlinear Flow in Random Heterogeneous Porous Media
A neural operator framework embeds a snapshot-derived multiscale basis as trunk to learn permeability-to-solution mappings for nonlinear flow, avoiding online coarse nonlinear solves.