Neural Network (Week 4) : Application

# Deep Neural Network for Image Classification: Application

# ## 1 - Packages
import time
import numpy as np
import h5py
import matplotlib.pyplot as plt
import scipy
from PIL import Image
from scipy import ndimage
from dnn_app_utils_v3 import *

# ## 2 - Dataset
train_x_orig, train_y, test_x_orig, test_y, classes = load_data()
index = 10
plt.imshow(train_x_orig[index])

m_train = train_x_orig.shape[0]
num_px = train_x_orig.shape[1]
m_test = test_x_orig.shape[0]


# Reshape the training and test examples
train_x_flatten = train_x_orig.reshape(train_x_orig.shape[0], -1).T
test_x_flatten = test_x_orig.reshape(test_x_orig.shape[0], -1).T


# Standardize data to have feature values between 0 and 1.
train_x = train_x_flatten/255.
test_x = test_x_flatten/255.

### CONSTANTS DEFINING THE MODEL ####
n_x = 12288 # num_px * num_px * 3
n_h = 7
n_y = 1
layers_dims = (n_x, n_h, n_y)


def two_layer_model(X, Y, layers_dims, learning_rate = 0.0075,
    num_iterations = 3000, print_cost=False):
    grads = {}
    costs = []
    m = X.shape[1]
    (n_x, n_h, n_y) = layers_dims
    
    parameters = initialize_parameters(n_x, n_h, n_y)
    
    W1 = parameters["W1"]
    b1 = parameters["b1"]
    W2 = parameters["W2"]
    b2 = parameters["b2"]
    
    # Loop (gradient descent)
    for i in range(0, num_iterations):
        A1, cache1 = linear_activation_forward(X, W1, b1, 'relu')
        A2, cache2 = linear_activation_forward(A1, W2, b2, 'sigmoid')
        
        cost = compute_cost(A2, Y)
        
        dA2 = - (np.divide(Y, A2) - np.divide(1 - Y, 1 - A2))

        dA1, dW2, db2 = linear_activation_backward(dA2, cache2, 'sigmoid')
        dA0, dW1, db1 = linear_activation_backward(dA1, cache1, 'relu')        

        grads['dW1'] = dW1
        grads['db1'] = db1
        grads['dW2'] = dW2
        grads['db2'] = db2
        
        parameters = update_parameters(parameters, grads, learning_rate)

        W1 = parameters["W1"]
        b1 = parameters["b1"]
        W2 = parameters["W2"]
        b2 = parameters["b2"]
        
        # Print the cost every 100 training example
        if print_cost and i % 100 == 0:
            print("Cost after iteration {}: {}".format(i, np.squeeze(cost)))
        if print_cost and i % 100 == 0:
            costs.append(cost)
       
    # plot the cost
    plt.plot(np.squeeze(costs))
    plt.ylabel('cost')
    plt.xlabel('iterations (per tens)')
    plt.title("Learning rate =" + str(learning_rate))
    plt.show()
    
    return parameters


parameters = two_layer_model(
    train_x, train_y, layers_dims = (n_x, n_h, n_y),
    num_iterations = 2500, print_cost=True)

predictions_train = predict(train_x, train_y, parameters)

predictions_test = predict(test_x, test_y, parameters)



layers_dims = [12288, 20, 7, 5, 1]

def L_layer_model(X, Y, layers_dims, learning_rate = 0.0075,
    num_iterations = 3000, print_cost=False):
    costs = []
    parameters = initialize_parameters_deep(layers_dims)
    
    # Loop (gradient descent)
    for i in range(0, num_iterations):
        AL, caches = L_model_forward(X, parameters)
        cost = compute_cost(AL, Y)
        grads = L_model_backward(AL, Y, caches)

        parameters = update_parameters(parameters, grads, learning_rate)
                
        # Print the cost every 100 training example
        if print_cost and i % 100 == 0:
            print ("Cost after iteration %i: %f" %(i, cost))
        if print_cost and i % 100 == 0:
            costs.append(cost)
            
    # plot the cost
    plt.plot(np.squeeze(costs))
    plt.ylabel('cost')
    plt.xlabel('iterations (per tens)')
    plt.title("Learning rate =" + str(learning_rate))
    plt.show()
    
    return parameters