def fit(self,
input_matrix: Union[sparse.csr_matrix, np.ndarray],
seeds: Union[np.ndarray, dict] = None,
seeds_row: Union[np.ndarray, dict] = None,
seeds_col: Union[np.ndarray, dict] = None) -> 'RankClassifier':
"""Fit algorithm to data.
Parameters
----------
input_matrix :
Adjacency matrix or biadjacency matrix of the graph.
seeds :
Seed nodes (labels as dictionary or array; negative values ignored).
seeds_row, seeds_col :
Seed rows and columns (for bipartite graphs).
Returns
-------
self: :class:`RankClassifier`
"""
adjacency, seeds_labels, bipartite = get_adjacency_seeds(
input_matrix,
seeds=seeds,
seeds_row=seeds_row,
seeds_col=seeds_col)
seeds_labels = seeds_labels.astype(int)
classes, n_classes = check_labels(seeds_labels)
seeds_all = self._process_seeds(seeds_labels)
local_function = partial(self.algorithm.fit_transform, adjacency)
with Pool(self.n_jobs) as pool:
scores = np.array(pool.map(local_function, seeds_all))
scores = scores.T
scores = self._process_scores(scores)
scores = normalize(scores)
membership = sparse.coo_matrix(scores)
membership.col = classes[membership.col]
labels = np.argmax(scores, axis=1)
self.labels_ = classes[labels]
self.membership_ = sparse.csr_matrix(membership,
shape=(adjacency.shape[0],
np.max(seeds_labels) + 1))
if bipartite:
self._split_vars(input_matrix.shape)
return self
def _process_seeds(labels_seeds: np.ndarray) -> list:
"""Make one-vs-all seed labels from seeds.
Parameters
----------
labels_seeds
Returns
-------
List of seeds vectors.
"""
seeds_all = []
classes, _ = check_labels(labels_seeds)
for label in classes:
seeds = np.array(labels_seeds == label).astype(int)
seeds_all.append(seeds)
return seeds_all
def fit(self, adjacency: Union[sparse.csr_matrix, np.ndarray],
seeds: Union[np.ndarray, dict]) -> 'RankClassifier':
"""Fit algorithm to the data.
Parameters
----------
adjacency:
Adjacency matrix of the graph.
seeds:
Seed nodes (labels as dictionary or array; negative values ignored).
Returns
-------
self: :class:`RankClassifier`
"""
n = adjacency.shape[0]
seeds_labels = check_seeds(seeds, n).astype(int)
classes, n_classes = check_labels(seeds_labels)
seeds_all = self._process_seeds(seeds_labels)
local_function = partial(self.algorithm.fit_transform, adjacency)
with Pool(self.n_jobs) as pool:
scores = np.array(pool.map(local_function, seeds_all))
scores = scores.T
scores = self._process_scores(scores)
scores = normalize(scores)
membership = sparse.coo_matrix(scores)
membership.col = classes[membership.col]
labels = np.argmax(scores, axis=1)
self.labels_ = classes[labels]
self.membership_ = sparse.csr_matrix(membership,
shape=(n,
np.max(seeds_labels) + 1))
return self
def hot_and_cold_seeds(labels_seeds):
"""Make one-vs-all seed labels from seeds.
Parameters
----------
labels_seeds :
Returns
-------
seeds_all: list
Personalization vectors.
"""
seeds_all = []
classes, _ = check_labels(labels_seeds)
for label in classes:
seeds = -np.ones_like(labels_seeds)
seeds[labels_seeds == label] = 1
ix = np.logical_and(labels_seeds != label, labels_seeds >= 0)
seeds[ix] = 0
seeds_all.append(seeds)
return seeds_all
def process_seeds(labels_seeds, temperature_max: float = 1):
"""Make one-vs-all seed labels from seeds.
Parameters
----------
labels_seeds :
temperature_max
Returns
-------
personalizations: list
Personalization vectors.
"""
personalizations = []
classes, _ = check_labels(labels_seeds)
for label in classes:
personalization = -np.ones(labels_seeds.shape[0])
personalization[labels_seeds == label] = temperature_max
ix = np.logical_and(labels_seeds != label, labels_seeds >= 0)
personalization[ix] = 0
personalizations.append(personalization)
return personalizations