Search papers, labs, and topics across Lattice.
4
0
5
13
Train sparse GNNs with dynamic graph topologies across epochs without re-extracting structure, achieving state-of-the-art performance even at extreme edge budgets.
A unified layer equation for GNNs reveals how over 200 architectures can be systematically compared and optimized for performance.
TopoFormer achieves state-of-the-art performance in graph learning by seamlessly integrating topological structures into attention-based models, outperforming traditional methods.
Transfer from node classification to link prediction is consistently advantageous, but the reverse can degrade performance unless structural conditions are ideal.
