Sampling Triangulations and Calabi-Yau Threefolds with Autoregressive GNNs
Researchers have developed `dualGNN`, an autoregressive Graph Neural Network designed to sample fine, regular triangulations of convex polytopes. This model operates on a generalized dual graph and is invariant to certain symmetries, ensuring uniform sampling for polygons. `dualGNN` is computationally efficient, requiring minimal training time and resources, and has been successfully applied to string theory to uniformly sample Calabi-Yau threefolds. AI
IMPACT Introduces a novel GNN architecture for complex geometric sampling, potentially accelerating research in theoretical physics.