def life_time(Zpy, ns):
"""
life-time criterion for automatic selection of the number of clusters
[porting from life-time implementation on matlab]
Input:
Zpy (array): input data array of shape (number samples x number features).
ns (int): number of samples.
Output:
res (dict): output dict with indexes for each cluster determined.
Example: res = {'-1': noise indexes list,
'0': cluster 0 indexes list,
'1': cluster 1 indexes list}
Configurable fields:{"name": "cluster.dbscan", "config": {"min_samples": "10", "eps": "0.95", "metric": "euclidean"}, "inputs": ["data"], "outputs": ["core_samples", "labels"]}
See Also:
Example:
References:
.. [1]
"""
Z = hierarchy.to_mlab_linkage(Zpy)
#dif=Z[1:,2]-Z[0:-1,2]
dif = np.diff(Z[:, 2])
indice = np.argmax(dif)
maximo = dif[indice]
indice = Z[find(Z[:, 2] > Z[indice, 2]), 2]
if indice == []:
cont = 1
else:
cont = len(indice) + 1
# th = maximo
#testing the situation when only 1 cluster is present
#max>2*min_interval -> nc=1
minimo = np.min(dif[pl.find(dif != 0)])
if minimo != maximo: #se maximo=minimo e' porque temos um matriz de assocs perfeita de 0s e 1s
if maximo < 2 * minimo:
cont = 1
nc_stable = cont
if nc_stable > 1:
labels = hierarchy.fcluster(hierarchy.from_mlab_linkage(Z), nc_stable,
'maxclust')
else: #ns_stable=1
labels = np.arange(ns, dtype="int")
return labels
def test_mlab_linkage_conversion_multiple_rows(self):
# Tests from/to_mlab_linkage on linkage array with multiple rows.
Zm = np.asarray([[3, 6, 138], [4, 5, 219], [1, 8, 255], [2, 9, 268],
[7, 10, 295]])
Z = np.array(
[[2., 5., 138., 2.], [3., 4., 219., 2.], [0., 7., 255., 3.],
[1., 8., 268., 4.], [6., 9., 295., 6.]],
dtype=np.double)
assert_equal(from_mlab_linkage(Zm), Z)
assert_equal(to_mlab_linkage(Z), Zm)
def test_mlab_linkage_conversion_multiple_rows(self):
# Tests from/to_mlab_linkage on linkage array with multiple rows.
Zm = np.asarray([[3, 6, 138], [4, 5, 219],
[1, 8, 255], [2, 9, 268], [7, 10, 295]])
Z = np.array([[2., 5., 138., 2.],
[3., 4., 219., 2.],
[0., 7., 255., 3.],
[1., 8., 268., 4.],
[6., 9., 295., 6.]],
dtype=np.double)
assert_equal(from_mlab_linkage(Zm), Z)
assert_equal(to_mlab_linkage(Z), Zm)
def test_mlab_linkage_conversion_multiple_rows(self):
# Tests from/to_mlab_linkage on linkage array with multiple rows.
Zm = np.asarray([[3, 6, 138], [4, 5, 219], [1, 8, 255], [2, 9, 268], [7, 10, 295]])
Z = np.array(
[
[2.0, 5.0, 138.0, 2.0],
[3.0, 4.0, 219.0, 2.0],
[0.0, 7.0, 255.0, 3.0],
[1.0, 8.0, 268.0, 4.0],
[6.0, 9.0, 295.0, 6.0],
],
dtype=np.double,
)
assert_equal(from_mlab_linkage(Zm), Z)
assert_equal(to_mlab_linkage(Z), Zm)
def test_mlab_linkage_conversion_single_row(self):
# Tests from/to_mlab_linkage on linkage array with single row.
Z = np.asarray([[0., 1., 3., 2.]])
Zm = [[1, 2, 3]]
assert_equal(from_mlab_linkage(Zm), Z)
assert_equal(to_mlab_linkage(Z), Zm)
def test_mlab_linkage_conversion_empty(self):
# Tests from/to_mlab_linkage on empty linkage array.
X = np.asarray([])
assert_equal(from_mlab_linkage([]), X)
assert_equal(to_mlab_linkage([]), X)
def show_dendrogram(Z, **kwargs):
from scipy.cluster.hierarchy import dendrogram, from_mlab_linkage
from matplotlib import pyplot as plt
dendrogram(from_mlab_linkage(Z), **kwargs)
plt.show()
def test_mlab_linkage_conversion_single_row(self):
# Tests from/to_mlab_linkage on linkage array with single row.
Z = np.asarray([[0., 1., 3., 2.]])
Zm = [[1, 2, 3]]
assert_equal(from_mlab_linkage(Zm), Z)
assert_equal(to_mlab_linkage(Z), Zm)
def test_mlab_linkage_conversion_empty(self):
# Tests from/to_mlab_linkage on empty linkage array.
X = np.asarray([])
assert_equal(from_mlab_linkage([]), X)
assert_equal(to_mlab_linkage([]), X)
import scipy.cluster.hierarchy as hcluster
import numpy as np
import matplotlib.pyplot as plt
chained_linkage = np.array([[1, 2, 1], [9, 3, 2], [10, 4, 3], [11, 5, 4], [12, 6, 5], [13, 7, 6], [14, 8, 7]])
chained_linkage = hcluster.from_mlab_linkage(chained_linkage)
perfect_linkage = hcluster.from_mlab_linkage(np.array([[1,2,1], [3,4,1], [5,6,1], [7,8,1], [9,10,2], [11,12,2], [13,14,3]]))
hcluster.dendrogram(perfect_linkage, color_threshold=0)
plt.savefig('perfect_linkage.pdf')
plt.close('all')
hcluster.dendrogram(chained_linkage, color_threshold=0)
plt.savefig('worst_case_linkage.pdf')
plt.close('all')
def check_linkage_q(self, method):
# Tests linkage(Y, method) on the Q data set.
Z = linkage(eo['Q-X'], method)
Zmlab = eo['linkage-Q-%s' % method]
expectedZ = from_mlab_linkage(Zmlab)
assert_allclose(Z, expectedZ, atol=1e-06)
def check_linkage_tdist(self, method):
# Tests linkage(Y, method) on the tdist data set.
Z = linkage(_ytdist, method)
Zmlab = eo['linkage-%s-tdist' % method]
expectedZ = from_mlab_linkage(Zmlab)
assert_allclose(Z, expectedZ, atol=1e-10)
