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.