CAESAR
A unified framework of foundation and generative models for efficient compression of scientific data
Area: Artificial intelligence
CASS member: RAPIDS
Description
CAESAR is an AI foundation model for spatio-temporal scientific data reduction that stands for Conditional AutoEncoder with Super-resolution for Augmented Reduction. It provides two baseline models, CAESAR-V and CAESAR-D. CAESAR-V is built on a standard variational autoencoder, which encodes data into a latent space and uses learned priors for compact, information-rich representations. CAESAR-D incorporates conditional diffusion with the autoencoder for enhanced reconstruction quality. Both models have been trained on a variety of scientific datasets for generalization. The evaluation shows great potential of foundation models for addressing the growing storage and transmission challenges in scientific computing.