def _sample_from_zeros(n: int, sparse: sp.csr_matrix) -> List[List[int]]:
"""
Sample n zeros from spacse matrix.
Parameters
----------
n : int
Number of samples to get from matrix.
sparse : sp.csr_matrix
Sparse matrix.
Returns
-------
List[List[int]]
List of 2-D indices of zeros.
"""
zeros = np.argwhere(np.logical_not(sparse.todense()))
ids = np.random.choice(range(len(zeros)), size=(n, ))
return zeros[ids].tolist()
def predict_proba(self, X: csr_matrix):
return self.clf.predict_proba(X.todense())
def predict(self, X: csr_matrix):
X = X.todense() # TensorFlow/Skflow doesn't support sparse matrices
return self.clf.predict(X)
def sparse_to_tensor(in_sparse: sparse.csr_matrix) -> torch.Tensor:
return torch.Tensor(in_sparse.todense())
def to_tensor(inp: csr_matrix) -> tf.Tensor:
return tf.convert_to_tensor(inp.todense(), dtype=TENSOR_OUTPUT_TYPE)
def predict_proba(self, X: csr_matrix):
return self.clf.predict_proba(X.todense())
def predict(self, X: csr_matrix):
X = X.todense() # TensorFlow/Skflow doesn't support sparse matrices
return self.clf.predict(X)
def compute_jaccard_distance_matrix(feature_vectors: csr_matrix):
return squareform(pdist(feature_vectors.todense(), metric='jaccard'))
