DL_Course_SamU

You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

3.2 KiB

Raw Blame History

In [27]:

import numpy as np

In [28]:

def softmax(s, y):
    l_i = -np.log(np.exp(s[y])/sum(np.exp(s)))
    return l_i



def svm(s, y):
    m = np.maximum(np.zeros_like(s), s - s[y] + 1)
    m[y] = 0   
    l_i = np.sum(m)
    return l_i

In [34]:

s = np.array([[10, -2, 3],
              [10, 9, 9],
              [10, -100, -100]])

y = 0

In [35]:

print("softmax:\n")
for i in range(s.shape[0]):
    print(softmax(s[i,:], y))

print("\nSVM:\n")
for i in range(s.shape[0]):
    print(svm(s[i,:], y))

softmax:

0.0009176050495943237
0.5514447139320511
-0.0

SVM:

0
0
0

In [37]:

s[0,0] += 10
s

Out[37]:

array([[  20,   -2,    3],
       [  10,    9,    9],
       [  10, -100, -100]])

In [38]:

print("softmax:\n")
for i in range(s.shape[0]):
    print(softmax(s[i,:], y))

print("\nSVM:\n")
for i in range(s.shape[0]):
    print(svm(s[i,:], y))

softmax:

4.167832299541146e-08
0.5514447139320511
-0.0

SVM:

0
0
0

In [ ]:

3.2 KiB Raw Blame History

3.2 KiB

Raw Blame History