def get_variances(self, one_variance=False):
"""Returns the squared variance for each cluster in a list of length K.
If one_variance is set to True, it will return the same variance for
each cluster as calculated by sigma^2 = d_max^2 / 2K
Otherwise, the variance will be determined individually using the
following formula:
sigma^2 = sum ||x - mu||^2 if the cluster contains more than
one input
sigma^2 = avg(all other sigma^2's) if the cluster contains only one
input
:param one_variance: Boolean of indicating how to calculate the
variances.
"""
if one_variance:
d_max = 0
for cluster in self.clusters:
this_d_max = max([utils.euclidean_distance(cluster.center,
x.center)
for x in self.clusters])
if this_d_max > d_max:
d_max = this_d_max
for cluster in self.clusters:
cluster.variance = (d_max / (2 * self.K)) ** 2
else:
one_input_clusters = []
for cluster in self.clusters:
if len(cluster.assigned_inputs) == 1:
one_input_clusters.append(cluster)
continue
cluster.variance = sum([utils.euclidean_distance(x,
cluster.center)
** 2 for x in cluster.assigned_inputs])
avg_variance = sum([cluster.variance for cluster in self.clusters])
avg_variance /= (len(self.clusters) - len(one_input_clusters))
for cluster in one_input_clusters:
cluster.variance = avg_variance
return [cluster.variance for cluster in self.clusters]
def distance_to_point(other_point):
return utils.euclidean_distance(sample, other_point)
