def eps_nn(X, eps=1.):
"""Returns the eps-nearest-neighbours graph of the data
Parameters
----------
X, array of shape (n_samples, n_features), input data
eps, float, optional: the neighborhood width
Returns
-------
the resulting graph instance
"""
from nipy.algorithms.routines.fast_distance import euclidean_distance
if np.size(X) == X.shape[0]:
X = np.reshape(X, (np.size(X), 1))
try:
eps = float(eps)
except:
"eps cannot be cast to a float"
if np.isnan(eps):
raise ValueError('eps is nan')
if np.isinf(eps):
raise ValueError('eps is inf')
dist = euclidean_distance(X)
dist = dist * (dist < eps)
# this would is just for numerical reasons
dist -= np.diag(np.diag(dist))
return wgraph_from_adjacency(dist)
def knn(X, k=1):
"""returns the k-nearest-neighbours graph of the data
Parameters
----------
X, array of shape (n_samples, n_features): the input data
k, int, optional: is the number of neighbours considered
Returns
-------
the corresponding WeightedGraph instance
Note
----
The knn system is symmeterized: if (ab) is one of the edges
then (ba) is also included
"""
from nipy.algorithms.routines.fast_distance import euclidean_distance
if np.size(X) == X.shape[0]:
X = np.reshape(X, (np.size(X), 1))
try:
k = int(k)
except:
"k cannot be cast to an int"
if np.isnan(k):
raise ValueError('k is nan')
if np.isinf(k):
raise ValueError('k is inf')
k = min(k, X.shape[0] - 1)
# create the distance matrix
dist = euclidean_distance(X)
sorted_dist = dist.copy()
sorted_dist.sort(0)
# neighbour system
bool_knn = dist < sorted_dist[k + 1]
bool_knn += bool_knn.T
bool_knn -= np.diag(np.diag(bool_knn))
dist *= (bool_knn > 0)
return wgraph_from_adjacency(dist)
