Transformer

Efficient adaptation of ROMs for unsteady flows using data assimilation

This paper proposes an efficient retraining strategy for parameterized reduced-order models, which adapts a variational autoencoder and transformer architecture to out-of-sample flow regimes from sparse observations.

Latent attention on masked patches for flow reconstruction (LAMP)

This paper introduces LAMP, an interpretable regression-based vision transformer that reconstructs full flow fields from heavily masked and noisy inputs, and yields optimal sensor-placement maps.