def cal_info(X, Y, k):
"""
calculate the max energy distance among random vector Xs that condition
on k classes, associated with the random discrete variable Y
Requirments:
dcor, which is on PyPi and can be installed using pip and conda
Input:
X (:obj:`numpy.ndarray`): matrix of size n x d, n observations of X
where X is a d-dimensional random vector taking continuous values
Y (:obj:`numpy.ndarray`): matrix of size n x 1, n observations of Y
where Y is a 1-dimensional random vector taking discrete values
0,1,...k-1
k (:obj:`int`): number of discrete values Y can take
Output:
dist (:obj:`float`): mean energy distance of k assumed conditioanl
distributions among Xs, the larger the more difference and the more dependence of Y
err (:obj:`bool`): True, if the output can be used,
and False otherwise
Example:
If the X is a 1000 dimensional input, Y is the 10-classification label
and if there are 2000 observations, then
X: float-valued matrix of size 2000 x 1000
Y: discrete-valued vector of size 2000 x 1
k: integer 10
The algo returns near-zero if X and Y are independent, and positive otherwise
"""
n, d = X.shape
class_ind = []
for i in range(k):
ind = np.where(Y == i)[0]
if len(ind) > 0:
class_ind.append(ind)
# some class may be unobserved, redefine k
##how to redefine?
k0 = len(class_ind)
if k0 < 2:
return 0, False
# use the largest class as reference for pairwise calculation
ref = np.argmax(np.array([len(class_ind[i]) for i in range(k0)]))
X1 = X[class_ind[ref], :]
dists = []
for j in range(k0):
if j != ref:
disttemp = dcor.energy_distance(X1, X[class_ind[j], :])
dists.append(disttemp)
dist = np.max(np.array(dists))
return dist, True
print(feat)
print("Pearson's = {}".format(
pearsonr(biochemistry_data[feat],
questionnaire_data["HYPERTENSION"])[0]))
print("Spearman's = {}".format(
spearmanr(biochemistry_data[feat],
questionnaire_data["HYPERTENSION"])[0]))
print("Kendall's Tau = {}\n".format(
kendalltau(biochemistry_data[feat],
questionnaire_data["HYPERTENSION"])[0]))
print("Distance Correlation = {}".format(
dcor.distance_correlation(biochemistry_data[feat],
questionnaire_data["HYPERTENSION"])))
print("Energy Distance = {}\n".format(
dcor.energy_distance(biochemistry_data[feat],
questionnaire_data["HYPERTENSION"])))
#Plot correlation matrix using Pearson's correlation measure
if False:
import seaborn as sns
cols = list(biochemistry_data.columns)
corr_matrix = np.corrcoef(biochemistry_data[cols].values.T)
print(corr_matrix)
plt.figure(1, figsize=(12, 18))
sns.set(font_scale=1.0)
heat_map = sns.heatmap(corr_matrix,
cbar=False,
annot=True,
square=True,
fmt='.2f',
annot_kws={'size': 10},
def energy_dist(self):
try:
return energy_distance(self.s1, self.s2)
except:
return 0.0
level = get_level_from_z(z)
for l in level:
print(l)
print('\n')
print("Training: ")
level = get_level_from_z(tz)
for l in level:
print(l)
print('\n\n')
print(len(zs), len(train_z))
print("Avg Train Density: ", train_d / len(train_z))
print("Avg Gen Density: ", gen_d / len(zs))
print("Avg Train NL: ", train_nl / len(train_z))
print("Avg Gen NL: ", gen_nl / len(zs))
print("Avg Plagiarism: ", tot_p / (len(train_z) * len(zs)))
print("ED: ", dcor.energy_distance(z_mets, t_mets))
sys.exit()
z = get_z_from_file('CV_chunk_100.txt')
print(nonlinearity(z))
for _ in range(10):
z = torch.DoubleTensor(1, nz).normal_(0, 1)
level = get_level_from_z(z)
for l in level:
print(l)
print('\n')
interpolate_chunks('CV_chunk_100.txt', 'CV_chunk_1000.txt')
z = get_z_from_file('smb_chunk_150.txt')
''' Created on 5 Nov 2019 @author: snake91 ''' import dcor import numpy as np X = np.random.normal(size=100) Y = np.random.normal(size=100) dcor.energy_distance(X, Y)