강의시간

강의내용

실습여부

1

-  Why CNNs, MLPs, RNNs are insufficient for non-Euclidean data

-  Neural network for graphs & sets

-  Euclidean transformations, invariance, and equivariance

2

-  Invariant geometric GNNs (SchNet, DimeNet, and SphereNet)

-  Simple equivariant geometric GNNs (EGNN and NequIP)

O

3

-  Local frame-based geometric GNNs (ClofNet and LEFTNet)

-  Frame averaging for geometric GNNs (Frame averaging, FAENet)

O

4

-  Steerable features, rreducible representations, Wigner-D matrix, spherical harmonics, Clebsch-Gordan tensor product

-  Steerable geometric GNNs (Tensor field network, SE(3)-Transformer)

5

- Steerable geometric GNNs (Equiformer, MACE, eSCN, EquifomerV2)

O