Understanding Logistic Regression

Posted April 3, 2019 by Rokas Balsys

##### Architecture of the learning algorithm:

In this part we'll design a simple algorithm to distinguish cat images from dog images. We will build a Logistic Regression, using a Neural Network mindset. Figure bellow explains why Logistic Regression is actually a very simple Neural Network (one neuron): ##### Parts of our algorithm:

Main steps we will use to build "Neural Network" are:

• Define the model structure (data shape).
• Initialize model's parameters.
• Learn the parameters for the model by minimizing the cost:
- Calculate current loss (forward propagation).
- Calculate current gradient (backward propagation).
• Use the learned parameters to make predictions (on the test set).
• Analyse the results and conclude.

We will build above parts separately and then we will integrate them into one function called model().

In our first tutorial part, we already wrote a sigmoid function, so I will just copy it from there:

def sigmoid(z):
s = 1/(1+np.exp(-z))
return s


##### Forward propagation:

First, weight and bias values are propagated forward through the model to arrive at a predicted output. At each neuron/node, the linear combination of the inputs is then multiplied by an activation function — the sigmoid function in our example. In this process weights and biases are propagated from inputs to output is called forward propagation. After arriving at the predicted output, the loss for the training example is calculated. Mathematical expression of forward propagation algorithm for one example: $$z^{(i)} = w^T x^{(i)} + b \tag{1}$$ $$\hat{y}^{(i)} = a^{(i)} = sigmoid(z^{(i)})\tag{2}$$ $$\mathcal{L}(a^{(i)}, y^{(i)}) = - y^{(i)} \log(a^{(i)}) - (1-y^{(i)} ) \log(1-a^{(i)})\tag{3}$$
Then cost is computed by summing over all training examples: $$J = \frac{1}{m} \sum_{i=1}^m \mathcal{L}(a^{(i)}, y^{(i)})\tag{4}$$
And our final forward propagation cost function will look like this: $$J = -\frac{1}{m}\sum_{i=1}^{m}y^{(i)}\log(a^{(i)})+(1-y^{(i)})\log(1-a^{(i)})\tag{5}$$

##### Backward propagation:

Back propagation is the process of calculating the partial derivatives from the loss function back to the inputs, we are updating the values of w and b that lead us to the minimum. It’s helpful writing out the partial derivatives starting from dA to see how to arrive at dw and db. Mathematical expression of backward propagation (calculating derivatives): $$\frac{\partial J}{\partial w} = \frac{1}{m}X(A-Y)^T\tag{6}$$ $$\frac{\partial J}{\partial b} = \frac{1}{m} \sum_{i=1}^m (a^{(i)}-y^{(i)})\tag{7}$$

##### Coding forward and backward propagation:

So we will implement function explained above, but first lets see what are the inputs and outputs:

Arguments:
w - weights, a numpy array of size (ROWS * COLS * CHANNELS, 1)
b - bias, a scalar
X - data of size (ROWS * COLS * CHANNELS, number of examples)
Y - true "label" vector (containing 0 if dog, 1 if cat) of size (1, number of examples)

Return:
cost - cost for logistic regression
dw - gradient of the loss with respect to w, the same shape as w
db - gradient of the loss with respect to b, the same shape as b

Here is the code we wrote in video tutorial:

def propagate(w, b, X, Y):
m = X.shape

# FORWARD PROPAGATION (FROM X TO COST)
z = np.dot(w.T, X)+b # tag 1
A = sigmoid(z) # tag 2
cost = (-np.sum(Y*np.log(A)+(1-Y)*np.log(1-A)))/m # tag 5

# BACKWARD PROPAGATION (TO FIND GRAD)
dw = (np.dot(X,(A-Y).T))/m # tag 6
db = np.average(A-Y) # tag 7

cost = np.squeeze(cost)
"db": db}



Lets test above function with sample data:

w = np.array([[1.],[2.]])
b = 4.
X = np.array([[5., 6., -7.],[8., 9., -10.]])
Y = np.array([[1,0,1]])

grads, cost = propagate(w, b, X, Y)
print ("dw = " + str(grads["dw"]))
print ("db = " + str(grads["db"]))
print ("cost = " + str(cost))


As a result you should get:

dw = [[4.33333333]
[6.33333333]]
db = 2.934645119504845e-11
cost = 16.999996678946573


##### Full tutorial code:
import os
import cv2
import numpy as np
import matplotlib.pyplot as plt
import scipy

ROWS = 64
COLS = 64
CHANNELS = 3

TRAIN_DIR = 'Train_data/'
TEST_DIR = 'Test_data/'

train_images = [TRAIN_DIR+i for i in os.listdir(TRAIN_DIR)]
test_images =  [TEST_DIR+i for i in os.listdir(TEST_DIR)]

return cv2.resize(img, (ROWS, COLS), interpolation=cv2.INTER_CUBIC)

def prepare_data(images):
m = len(images)
X = np.zeros((m, ROWS, COLS, CHANNELS), dtype=np.uint8)
y = np.zeros((1, m))
for i, image_file in enumerate(images):
if 'dog' in image_file.lower():
y[0, i] = 1
elif 'cat' in image_file.lower():
y[0, i] = 0
return X, y

train_set_x, train_set_y = prepare_data(train_images)
test_set_x, test_set_y = prepare_data(test_images)

train_set_x_flatten = train_set_x.reshape(train_set_x.shape, ROWS*COLS*CHANNELS).T
test_set_x_flatten = test_set_x.reshape(test_set_x.shape, -1).T
'''
print("train_set_x shape " + str(train_set_x.shape))
print("train_set_x_flatten shape: " + str(train_set_x_flatten.shape))
print("train_set_y shape: " + str(train_set_y.shape))
print("test_set_x shape " + str(test_set_x.shape))
print("test_set_x_flatten shape: " + str(test_set_x_flatten.shape))
print("test_set_y shape: " + str(test_set_y.shape))
'''
train_set_x = train_set_x_flatten/255
test_set_x = test_set_x_flatten/255

def sigmoid(z):
s = 1/(1+np.exp(-z))
return s

def propagate(w, b, X, Y):
m = X.shape

# FORWARD PROPAGATION (FROM X TO COST)
z = np.dot(w.T, X)+b
A = sigmoid(z)
cost = (-np.sum(Y*np.log(A)+(1-Y)*np.log(1-A)))/m

# BACKWARD PROPAGATION (TO FIND GRAD)
dw = (np.dot(X,(A-Y).T))/m
db = np.average(A-Y)

cost = np.squeeze(cost)
"db": db}

w = np.array([[1.],[2.]])
b = 4.
X = np.array([[5., 6., -7.],[8., 9., -10.]])
Y = np.array([[1,0,1]])

grads, cost = propagate(w, b, X, Y)
print ("dw = " + str(grads["dw"]))
print ("db = " + str(grads["db"]))
print ("cost = " + str(cost))


So in this tutorial we defined general learning architecture and defined steps needed to implement learning model. I explained what is forward/backward propagation and we learned how to implement them in code. In next tutorial we will continue with optimization algorithm.