常用激活函数及偏导_激活 函数 偏导

常用激活函数及偏导

derivative.py

import numpy as np
import matplotlib.pyplot as plt

plt.subplots_adjust(hspace=0.5 , wspace=0.5)
rows = 3
cols = 2

def plot_style(ax):
    # 设置轴隐藏
    ax.spines['top'].set_visible(False) 
    ax.spines['right'].set_visible(False)
    # 设置轴位置
    ax.spines['bottom'].set_position(('axes', 0.5))
    ax.spines['left'].set_position(('axes', 0.5))
    # 设置刻度间隔
    # ax.xaxis.set_major_locator(plt.MultipleLocator(1))
    # ax.yaxis.set_major_locator(plt.MultipleLocator(1))
    # 设置刻度范围
    plt.xlim(-10,10)
    plt.ylim(-2,2)

# --------------------------------------

x=np.linspace(-10,10,100)

def Sigmoid(x) :
    a = 1/(1+np.exp(-x))
    return a

def d_Sigmoid(x):
    a = Sigmoid(x)*(1-Sigmoid(x))
    return a

graph = plt.subplot(rows,cols,1)
plot_style(graph)
plt.plot(x,Sigmoid(x),label="σ(x)")
plt.plot(x,d_Sigmoid(x),label="σ`(x)")
# plt.title('σ(x) & σ`(x)=σ(1-σ)')
plt.legend()

# plt.show()

# --------------------------------------
def LeakyReLU(x,p):
    cond = x>=0
    a = np.where(cond,x,p*x) 
    return a 
def d_LeakyReLU(x,p):
    cond = x>=0
    a = np.where(cond,1,p) 
    return a 

x=np.linspace(-10,10,100)
graph = plt.subplot(rows,cols,2)
plot_style(graph)
p = 0.2
plt.plot(x,LeakyReLU(x,p),label='LeakyReLU(x)')
plt.plot(x,d_LeakyReLU(x,p),label='`LeakyReLU(x)')
plt.title(f'p={p}')
plt.legend()

# --------------------------------------
def ReLU(x) :
    a = LeakyReLU(x,0)
    return a 

def d_ReLU(x) :
    a = np.array(x,copy=True)
    a[x<0] = 0
    a[x>=0] = 1
    return a 

x=np.linspace(-10,10,100)
graph = plt.subplot(rows,cols,3)
plot_style(graph)
plt.plot(x,ReLU(x),label='ReLU(x)')
plt.plot(x,d_ReLU(x),label='`ReLU(x)')
# plt.title('ReLU(x) & `ReLU(x)')
plt.legend()
# --------------------------------------


def tanh(x):
    a = 2*Sigmoid(2*x) - 1
    return a

def d_tanh(x):
    b = np.power(tanh(x),2) 
    a = 1 - b
    return a

x=np.linspace(-10,10,100)
graph = plt.subplot(rows,cols,4)
plot_style(graph)
plt.plot(x,tanh(x),label='tanh(x)')
plt.plot(x,d_tanh(x),label='`tanh(x)')
# plt.title(' tanh(x) & `tanh(x) ')
plt.legend()
# --------------------------------------

# x=np.linspace(0,10,3)
x = [2,3,4]
def softmax(x):
    expx = np.exp(x)
    a = expx / np.sum(expx)
    return a

def d_softmax(x):
    p = softmax(x)
    count = len(x)
    out = []
    for i in range(count) :
        for j in range(count) :
            ret = 0
            if j==i:
                ret = p[i]*(1-p[j])
                print(f'dp[{i}]/dz[{j}] = p[{i}]*(1-p[{j})] = {ret}')
            else :
                ret = -p[i]*p[j]
                print(f'dp[{i}]/dz[{j}] = -p[{i}]*p[{j}] = {ret}')
            out.append(ret)
    return out

plt.subplot(rows,cols,5)
plt.scatter(x,softmax(x),label=f'softmax(zi)')
d_out = d_softmax(x)
len = len(x)
for i in range(len):
    plt.scatter(x,d_out[len*i:len*(i+1)],label=f'`softmax(z{i})')

plt.legend(loc='upper left', bbox_to_anchor=(1, 1))

# --------------------------------------

plt.show()

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137