|
forward, invert, evals = transformer(adj.to_dense()) |
The adj which is row normalized adjacency matrix is not symmetric, which is then passed to np.linalg.eigh for eigendecomposition, but this function only accepts a complex Hermitian (conjugate symmetric) or a real symmetric matrix. Do I miss something?
gfnn/raw_feats_citation.py
Line 55 in 81878f6
The adj which is row normalized adjacency matrix is not symmetric, which is then passed to
np.linalg.eighfor eigendecomposition, but this function only accepts a complex Hermitian (conjugate symmetric) or a real symmetric matrix. Do I miss something?