Search papers, labs, and topics across Lattice.
2
0
3
2
SPD-SheafNets learn richer geometric representations than standard GNNs by operating directly on matrix-valued features, achieving SOTA on molecular property prediction by capturing relationships between directions.
SPD Learn streamlines geometric deep learning with SPD matrices by providing a unified Python package that ensures manifold constraints via trivialization, enabling standard backpropagation for neural decoding.
