def testNearest2(self):
with self.test_session():
[indices, distances] = clustering_ops.nearest_neighbors(self._points,
self._centers, 2)
self.assertAllClose(indices.eval(), [[0, 1], [0, 1], [1, 0], [4, 3]])
self.assertAllClose(distances.eval(),
[[0., 2.], [5., 5.], [1., 5.], [0., 2.]])
def predict(self, X, with_distances=False):
""" predict.
Predict the closest cluster.
Arguments:
X: `1-D Array` or `2-D Array` of shape (n_samples, n_features).
The sample(s) to predict.
Return:
cluster_indices or (cluster_indices, distances).
"""
X, orig_ndim = prepare_X(X, 2, max_dim=2, min_dim=1, debug_msg="X")
with self.graph.as_default():
# Build Tree Graph
self._build_estimator()
if not self._pred.is_ready:
input = tf.placeholder(tf.float32, name='pred_input',
shape=[None, self.num_features])
output = c_ops.nearest_neighbors(
input, self._cluster_centers_vars, k=1)
self._build_pred(input, output)
indices, distances = self.session.run(self._pred.output_tensor,
feed_dict={self._pred.input_tensor: X})
indices = indices[0]
distances = distances[0]
if orig_ndim == 1:
indices = indices[0]
distances = distances[0]
if with_distances:
return indices, distances
return indices
def testNearest1(self):
with self.cached_session():
[indices, distances
] = clustering_ops.nearest_neighbors(self._points, self._centers,
1)
self.assertAllClose(indices.eval(), [[0], [0], [1], [4]])
self.assertAllClose(distances.eval(), [[0.], [5.], [1.], [0.]])
def testNearest5(self):
with self.test_session():
[indices, distances] = clustering_ops.nearest_neighbors(self._points,
self._centers, 5)
self.assertAllClose(indices.eval(),
self._expected_nearest_neighbor_indices[:, 0:5])
self.assertAllClose(
distances.eval(),
self._expected_nearest_neighbor_squared_distances[:, 0:5])
def predict(self, X, with_distances=False):
""" predict.
Predict the closest cluster.
Arguments:
X: `1-D Array` or `2-D Array` of shape (n_samples, n_features).
The sample(s) to predict.
Return:
cluster_indices or (cluster_indices, distances).
"""
X, orig_ndim = prepare_X(X, 2, max_dim=2, min_dim=1, debug_msg="X")
with self.graph.as_default():
# Build Tree Graph
self._build_estimator()
if not self._pred.is_ready:
input = tf.placeholder(tf.float32,
name='pred_input',
shape=[None, self.num_features])
output = c_ops.nearest_neighbors(input,
self._cluster_centers_vars,
k=1)
self._build_pred(input, output)
indices, distances = self.session.run(
self._pred.output_tensor,
feed_dict={self._pred.input_tensor: X})
indices = indices[0]
distances = distances[0]
if orig_ndim == 1:
indices = indices[0]
distances = distances[0]
if with_distances:
return indices, distances
return indices
def testNearest1(self):
with self.cached_session():
[indices, distances] = clustering_ops.nearest_neighbors(self._points,
self._centers, 1)
self.assertAllClose(indices.eval(), [[0], [0], [1], [4]])
self.assertAllClose(distances.eval(), [[0.], [5.], [1.], [0.]])
