Low-Rank Templates
Low-rank score templates allow diffuse motifs rather than rigid block-constant communities. Each template is parameterized by an orthonormal factor matrix, signed spectral weights, and a global scale. Positive and negative spectral weights…
1 sources - 4 claims
Low-rank score templates allow diffuse motifs rather than rigid block-constant communities. Each template is parameterized by an orthonormal factor matrix, signed spectral weights, and a global scale. Positive and negative spectral weights allow assortative and disassortative patterns. Inference focuses on reconstructed off-diagonal templates because only off-diagonal edges enter the likelihood.