def su_calculation(f1, f2):
"""
This function calculates the symmetrical uncertainty, where su(f1,f2) = 2*IG(f1,f2)/(H(f1)+H(f2))
Input
-----
f1: {numpy array}, shape (n_samples,)
f2: {numpy array}, shape (n_samples,)
Output
------
su: {float}
su is the symmetrical uncertainty of f1 and f2
"""
# calculate information gain of f1 and f2, t1 = ig(f1,f2)
t1 = information_gain(f1, f2)
# calculate entropy of f1, t2 = H(f1)
t2 = ee.entropyd(f1)
# calculate entropy of f2, t3 = H(f2)
t3 = ee.entropyd(f2)
# su(f1,f2) = 2*t1/(t2+t3)
su = 2.0*t1/(t2+t3)
return su
def su_calculation(f1, f2):
"""
This function calculates the symmetrical uncertainty, where su(f1,f2) = 2*IG(f1,f2)/(H(f1)+H(f2))
Input
-----
f1: {numpy array}, shape (n_samples,)
f2: {numpy array}, shape (n_samples,)
Output
------
su: {float}
su is the symmetrical uncertainty of f1 and f2
"""
# calculate information gain of f1 and f2, t1 = ig(f1,f2)
t1 = information_gain(f1, f2)
# calculate entropy of f1, t2 = H(f1)
t2 = ee.entropyd(f1)
# calculate entropy of f2, t3 = H(f2)
t3 = ee.entropyd(f2)
# su(f1,f2) = 2*t1/(t2+t3)
su = 2.0 * t1 / (t2 + t3 + sys.float_info.epsilon)
return su
def su_calculation(f1, f2):
"""
This function calculates the symmetrical uncertainty, where su(f1,f2) = 2*IG(f1,f2)/(H(f1)+H(f2))
Input
-----
f1: {numpy array}, shape (n_samples,)
f2: {numpy array}, shape (n_samples,)
Output
------
su: {float}
su is the symmetrical uncertainty of f1 and f2
"""
# calculate information gain of f1 and f2, t1 = ig(f1,f2)
t1 = information_gain(f1, f2)
# calculate entropy of f1, t2 = H(f1)
t2 = ee.entropyd(f1)
# calculate entropy of f2, t3 = H(f2)
t3 = ee.entropyd(f2)
# su(f1,f2) = 2*t1/(t2+t3)
if (t2 + t3 == 0): # avoid a division by zero error when t2+t3=0
su = 0
else:
su = 2.0 * t1 / (t2 + t3)
return su
def information_gain(f1, f2):
"""
This function calculates the information gain, where ig(f1,f2) = H(f1) - H(f1|f2)
Input
-----
f1: {numpy array}, shape (n_samples,)
f2: {numpy array}, shape (n_samples,)
Output
------
ig: {float}
"""
ig = ee.entropyd(f1) - conditional_entropy(f1, f2)
return ig
def information_gain(f1, f2):
"""
This function calculates the information gain, where ig(f1,f2) = H(f1) - H(f1|f2)
Input
-----
f1: {numpy array}, shape (n_samples,)
f2: {numpy array}, shape (n_samples,)
Output
------
ig: {float}
"""
ig = ee.entropyd(f1) - conditional_entropy(f1, f2)
return ig
def conditional_entropy(f1, f2):
"""
This function calculates the conditional entropy, where ce = H(f1) - I(f1;f2)
Input
-----
f1: {numpy array}, shape (n_samples,)
f2: {numpy array}, shape (n_samples,)
Output
------
ce: {float}
ce is conditional entropy of f1 and f2
"""
ce = ee.entropyd(f1) - ee.midd(f1, f2)
return ce
def conditional_entropy(f1, f2):
"""
This function calculates the conditional entropy, where ce = H(f1) - I(f1;f2)
Input
-----
f1: {numpy array}, shape (n_samples,)
f2: {numpy array}, shape (n_samples,)
Output
------
ce: {float}
ce is conditional entropy of f1 and f2
"""
ce = ee.entropyd(f1) - ee.midd(f1, f2)
return ce