def score_conditional_entropy(km):
"""UNIMPLEMENTED cluster quality measured by conditional entropy"""
raise NotImplemented
def score_within_cluster_distance(km):
"""UNIMPLEMENTED weighted average within-cluster pairwise distance"""
raise NotImplemented
score_within_cluster_distance.minimize = True
def score_between_cluster_distance(km):
"""Sum of distances from elements to 'nearest miss' centroids"""
return sum(min(km.distance(c, d) for j,c in enumerate(km.centroids) if j!=km.clusters[i]) for i,d in enumerate(km.data))
from Orange.misc import deprecated_function_name
score_betweenClusterDistance = deprecated_function_name(score_between_cluster_distance)
def score_silhouette(km, index=None):
"""Returns an average silhouette score of data instances.
:param km: a k-means clustering object.
:type km: :class:`KMeans`
:param index: if given, the functon returns just the silhouette score of that particular data instance.
:type index: integer
"""
if index == None:
return avg([score_silhouette(km, i) for i in range(len(km.data))])
cind = km.clusters[index]
a = avg([km.distance(km.data[index], ex) for i, ex in enumerate(km.data) if
the same experiment always gives the same results, despite
pseudo-randomness.random seed.
:type seed: int
:param call_compare_on_1st: If set, :obj:`BestOnTheFly` will suppose
that the candidates are lists are tuples, and it will call compare
with the first element of the tuple.
:type call_compare_on_1st: bool
:rtype: object
"""
bs=BestOnTheFly(compare, seed, call_compare_on_1st)
for i in x:
bs.candidate(i)
return bs.winner()
selectBest = deprecated_function_name(select_best)
@deprecated_keywords({"callCompareOn1st": "call_compare_on_1st"})
def select_best_index(x, compare=cmp, seed = 0, call_compare_on_1st = False):
"""Similar to :obj:`selectBest` except that it doesn't return the best object
but its index in the list x.
"""
bs=BestOnTheFly(compare, seed, call_compare_on_1st)
for i in x:
bs.candidate(i)
return bs.winner_index()
selectBestIndex = deprecated_function_name(select_best_index)
raise NotImplemented
score_within_cluster_distance.minimize = True
def score_between_cluster_distance(km):
"""Sum of distances from elements to 'nearest miss' centroids"""
return sum(
min(
km.distance(c, d) for j, c in enumerate(km.centroids)
if j != km.clusters[i]) for i, d in enumerate(km.data))
from Orange.misc import deprecated_function_name
score_betweenClusterDistance = deprecated_function_name(
score_between_cluster_distance)
def score_silhouette(km, index=None):
"""Returns an average silhouette score of data instances.
:param km: a k-means clustering object.
:type km: :class:`KMeans`
:param index: if given, the functon returns just the silhouette score of that particular data instance.
:type index: integer
"""
if index == None:
return avg([score_silhouette(km, i) for i in range(len(km.data))])
cind = km.clusters[index]
pseudo-randomness.random seed.
:type seed: int
:param call_compare_on_1st: If set, :obj:`BestOnTheFly` will suppose
that the candidates are lists are tuples, and it will call compare
with the first element of the tuple.
:type call_compare_on_1st: bool
:rtype: object
"""
bs = BestOnTheFly(compare, seed, call_compare_on_1st)
for i in x:
bs.candidate(i)
return bs.winner()
selectBest = deprecated_function_name(select_best)
@deprecated_keywords({"callCompareOn1st": "call_compare_on_1st"})
def select_best_index(x, compare=cmp, seed=0, call_compare_on_1st=False):
"""Similar to :obj:`selectBest` except that it doesn't return the best object
but its index in the list x.
"""
bs = BestOnTheFly(compare, seed, call_compare_on_1st)
for i in x:
bs.candidate(i)
return bs.winner_index()
selectBestIndex = deprecated_function_name(select_best_index)
