-
Notifications
You must be signed in to change notification settings - Fork 0
/
auxiliary.py
73 lines (58 loc) · 2.53 KB
/
auxiliary.py
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
import numpy as np
def relu(Z):
# Rectified Linear Unit Function used for the traversal of the Neural Network
return np.maximum(0, Z)
def relu_derivative(Z):
# Derivative of the ReLU function (useful for back propagation)
Z = np.where(Z <= 0, 0, Z)
Z = np.where(Z > 0, 1, Z)
return Z
def leaky_relu(Z):
# Leaky Rectified Linear Unit Function
c = 0.001
Z = np.where(Z <= 0, Z * c, Z)
Z = np.where(Z > 0, 1, Z)
return Z
def sigmoid(Z):
# Sigmoid (Logistic) Function used for the last step of the forward propagation phase
return 1 / (1 + np.exp(-Z))
def sigmoid_derivative(Z):
# Derivative of the Sigmoid (Logistic) Function (useful for back propagation)
return sigmoid(Z) * (1 - sigmoid(Z))
def logarithm(Z):
# Logarithm function that does not allow 0 values (because log(0) = -infinity)
constant = 0.000001
Z = np.where(Z == 0.0, constant, Z)
return np.log(Z)
def tanh(Z):
# Hyperbolic Tangent function - alternative for the Sigmoid Function
return (np.exp(Z) - np.exp(-Z)) / (np.exp(Z) + np.exp(-Z))
def tanh_derivative(Z):
# Hyperbolic Tangent Derivative
return 1 / np.power(np.cosh(Z), 2)
def softmax(Z):
# Softmax Function
return np.exp(Z) / np.sum(np.exp(Z), axis=0)
def dictionary_to_vector(dict):
# Function that given a dictionary will output a vector with the values from the dictionary
vec = []
for i in dict.keys():
vec.append(dict[i])
return vec
def vector_to_dictionary(vec):
# Function that given a vector will output a dictionary
dic = {}
for i in range(0, len(vec)-1, 2):
dic['W' + str(int(i/2) + 1)] = vec[i]
dic['b' + str(int(i/2) + 1)] = vec[i+1]
return dic
def initialize_adam(parameters, dimensions):
# Function used to initialise the parameters needed for Adam optimization
v = {} # Initialise empty dictionary for v values
s = {} # Initialise empty dictionary for s values
for l in range(len(dimensions) - 1):
v["dW" + str(l + 1)] = np.zeros((parameters["W" + str(l + 1)].shape[0], parameters["W" + str(l + 1)].shape[1]))
v["db" + str(l + 1)] = np.zeros((parameters["b" + str(l + 1)].shape[0], parameters["b" + str(l + 1)].shape[1]))
s["dW" + str(l + 1)] = np.zeros((parameters["W" + str(l + 1)].shape[0], parameters["W" + str(l + 1)].shape[1]))
s["db" + str(l + 1)] = np.zeros((parameters["b" + str(l + 1)].shape[0], parameters["b" + str(l + 1)].shape[1]))
return v, s