CoCalc -- lab_utils

GitHub Repository: greyhatguy007/machine-learning-specialization-coursera
Path: blob/main/C1 - Supervised Machine Learning - Regression and Classification/week1/Optional Labs/lab_utils_common.py
²²⁰¹ views
1
""" 
2
lab_utils_common.py
3
    functions common to all optional labs, Course 1, Week 2 
4
"""
5

6
import numpy as np
7
import matplotlib.pyplot as plt
8

9
plt.style.use('./deeplearning.mplstyle')
10
dlblue = '#0096ff'; dlorange = '#FF9300'; dldarkred='#C00000'; dlmagenta='#FF40FF'; dlpurple='#7030A0';
11
dlcolors = [dlblue, dlorange, dldarkred, dlmagenta, dlpurple]
12
dlc = dict(dlblue = '#0096ff', dlorange = '#FF9300', dldarkred='#C00000', dlmagenta='#FF40FF', dlpurple='#7030A0')
13

14

15
##########################################################
16
# Regression Routines
17
##########################################################
18

19
#Function to calculate the cost
20
def compute_cost_matrix(X, y, w, b, verbose=False):
21
    """
22
    Computes the gradient for linear regression
23
     Args:
24
      X (ndarray (m,n)): Data, m examples with n features
25
      y (ndarray (m,)) : target values
26
      w (ndarray (n,)) : model parameters  
27
      b (scalar)       : model parameter
28
      verbose : (Boolean) If true, print out intermediate value f_wb
29
    Returns
30
      cost: (scalar)
31
    """
32
    m = X.shape[0]
33

34
    # calculate f_wb for all examples.
35
    f_wb = X @ w + b
36
    # calculate cost
37
    total_cost = (1/(2*m)) * np.sum((f_wb-y)**2)
38

39
    if verbose: print("f_wb:")
40
    if verbose: print(f_wb)
41

42
    return total_cost
43

44
def compute_gradient_matrix(X, y, w, b):
45
    """
46
    Computes the gradient for linear regression
47

48
    Args:
49
      X (ndarray (m,n)): Data, m examples with n features
50
      y (ndarray (m,)) : target values
51
      w (ndarray (n,)) : model parameters  
52
      b (scalar)       : model parameter
53
    Returns
54
      dj_dw (ndarray (n,1)): The gradient of the cost w.r.t. the parameters w.
55
      dj_db (scalar):        The gradient of the cost w.r.t. the parameter b.
56

57
    """
58
    m,n = X.shape
59
    f_wb = X @ w + b
60
    e   = f_wb - y
61
    dj_dw  = (1/m) * (X.T @ e)
62
    dj_db  = (1/m) * np.sum(e)
63

64
    return dj_db,dj_dw
65

66

67
# Loop version of multi-variable compute_cost
68
def compute_cost(X, y, w, b):
69
    """
70
    compute cost
71
    Args:
72
      X (ndarray (m,n)): Data, m examples with n features
73
      y (ndarray (m,)) : target values
74
      w (ndarray (n,)) : model parameters  
75
      b (scalar)       : model parameter
76
    Returns
77
      cost (scalar)    : cost
78
    """
79
    m = X.shape[0]
80
    cost = 0.0
81
    for i in range(m):
82
        f_wb_i = np.dot(X[i],w) + b           #(n,)(n,)=scalar
83
        cost = cost + (f_wb_i - y[i])**2
84
    cost = cost/(2*m)
85
    return cost 
86

87
def compute_gradient(X, y, w, b):
88
    """
89
    Computes the gradient for linear regression
90
    Args:
91
      X (ndarray (m,n)): Data, m examples with n features
92
      y (ndarray (m,)) : target values
93
      w (ndarray (n,)) : model parameters  
94
      b (scalar)       : model parameter
95
    Returns
96
      dj_dw (ndarray Shape (n,)): The gradient of the cost w.r.t. the parameters w.
97
      dj_db (scalar):             The gradient of the cost w.r.t. the parameter b.
98
    """
99
    m,n = X.shape           #(number of examples, number of features)
100
    dj_dw = np.zeros((n,))
101
    dj_db = 0.
102

103
    for i in range(m):
104
        err = (np.dot(X[i], w) + b) - y[i]
105
        for j in range(n):
106
            dj_dw[j] = dj_dw[j] + err * X[i,j]
107
        dj_db = dj_db + err
108
    dj_dw = dj_dw/m
109
    dj_db = dj_db/m
110

111
    return dj_db,dj_dw
112

113

114
Product

Resources

Company
