def features_sparse_modular (A): from scipy.sparse import csc_matrix from modshogun import SparseRealFeatures from numpy import array, float64, all # sparse representation X of dense matrix A # note, will work with types other than float64 too, # but requires recent scipy.sparse X=csc_matrix(A) #print(A) # create sparse shogun features from dense matrix A a=SparseRealFeatures(A) a_out=a.get_full_feature_matrix() #print(a_out) assert(all(a_out==A)) #print(a_out) # create sparse shogun features from sparse matrix X a.set_sparse_feature_matrix(X) a_out=a.get_full_feature_matrix() #print(a_out) assert(all(a_out==A)) # create sparse shogun features from sparse matrix X a=SparseRealFeatures(X) a_out=a.get_full_feature_matrix() #print(a_out) assert(all(a_out==A)) # obtain (data,row,indptr) csc arrays of sparse shogun features z=csc_matrix(a.get_sparse_feature_matrix()) z_out=z.todense() #print(z_out) assert(all(z_out==A))
