past.Utils.spatial_prior_graph¶
- past.Utils.spatial_prior_graph(feature_matrix, k_neighbors, v=1)¶
Construct spatial prior graph for metric learning
- Parameters:
feature_matrix – spatial corrdinate matrix
k_neighbors – number of neighbors to construct graph
v – scale factor in student’s t kernel
- Returns:
spatial graph of sparse matrix format
- Return type:
scipy.sparse.csr_matrx