CSE572 Statistical Machine Learning - Project 3 - Classification Using Neural Networks and Deep Learning Project

import warnings
warnings.filterwarnings("ignore")
import keras
from tensorflow.keras.utils import to_categorical
from keras.datasets import mnist
from keras.models import Sequential
from keras.layers import Dense, Dropout, Flatten
from keras.layers import Conv2D, MaxPooling2D
from keras import backend as K
import tensorflow as tf

from sklearn.metrics import classification_report
from sklearn.metrics import confusion_matrix, ConfusionMatrixDisplay

import os
import multiprocessing

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt

# Number of logical CPUs
num_cores = multiprocessing.cpu_count()
print(f"Number of CPU cores available: {num_cores}")

# Number of GPUs
print("Num GPUs Available:", len(tf.config.list_physical_devices('GPU')))

batch_size = 64

num_classes = 10

epochs = 12

# input image dimensions
img_rows, img_cols = 28, 28

kernel_size = (5,5)

pool_size = (2,2)

# the data, split between train and test sets
(x_train, y_train), (x_test, y_test)=tf.keras.datasets.mnist.load_data('mnist.npz')

x_train.shape

y_train.shape

x_test.shape

y_test.shape

ticks = np.arange(0, 28+2, 2)
#ticks

plt.imshow(x_train[0], cmap='grey')
plt.xlim(0, 28)
plt.xticks(ticks=ticks)
plt.yticks(ticks=ticks)
plt.show()

# Display the first 10 images
num_images_to_display = 10
plt.figure(figsize=(10, 5))  # Adjust the figure size

for i in range(num_images_to_display):
    plt.subplot(2, 5, i + 1)  # Create a 2x5 grid for 10 images
    plt.imshow(x_train[i], cmap='gray')  # Display each image
    plt.title(f'Train Image {i + 1}')
    plt.axis('off')  # Hide axes for better visualization

plt.tight_layout()
plt.show()

y_train[0:10]

# Display the first 10 images
num_images_to_display = 10
plt.figure(figsize=(10, 5))  # Adjust the figure size

for i in range(num_images_to_display):
    plt.subplot(2, 5, i + 1)  # Create a 2x5 grid for 10 images
    plt.imshow(x_test[i], cmap='gray')  # Display each image
    plt.title(f'Test Image {i + 1}')
    plt.axis('off')  # Hide axes for better visualization

plt.tight_layout()
plt.show()

y_test[0:10]

if K.image_data_format() == 'channels_first':
    x_train = x_train.reshape(x_train.shape[0], 1, img_rows, img_cols)
    x_test = x_test.reshape(x_test.shape[0], 1, img_rows, img_cols)
    input_shape = (1, img_rows, img_cols)
else:
    x_train = x_train.reshape(x_train.shape[0], img_rows, img_cols, 1)
    x_test = x_test.reshape(x_test.shape[0], img_rows, img_cols, 1)
    input_shape = (img_rows, img_cols, 1)

x_train.shape

x_test.shape

x_train = x_train.astype('float32')
x_test = x_test.astype('float32')
x_train /= 255
x_test /= 255

print('x_train shape:', x_train.shape)
print(x_train.shape[0], 'train samples')
print(x_test.shape[0], 'test samples')

# convert class vectors to binary class matrices
y_train = to_categorical(y_train, num_classes)
y_test = to_categorical(y_test, num_classes)

y_train.shape

y_train[0:5]

y_test.shape

y_test[0:5]

# Artificial Neural Network
model = Sequential()

model

# First layer is a convolutional layer with 6 filters, each of size 3x3
model.add(Conv2D(
    filters=6, 
    kernel_size=kernel_size,
    strides=(1, 1),
    activation='relu',
    input_shape=input_shape))

# Second layer is a max pooling layer with a pool size of 2x2
model.add(MaxPooling2D(
    pool_size=pool_size,
    strides=(1, 1)))

# Third layer is a convolutional layer with 16 filters, each of size 3x3
model.add(Conv2D(
    filters=16, 
    kernel_size=kernel_size,
    strides=(1, 1),
    activation='relu'))

# Fourth layer is a max pooling layer with a pool size of 2x2
model.add(MaxPooling2D(
    pool_size=pool_size,
    strides=(1, 1)))

# Fifth layer is a flatten layer
model.add(Flatten())

# Sixth layer is a dense layer with 120 units
model.add(Dense(120, activation='relu'))

# Seventh layer is a dense layer with 84 units
model.add(Dense(84, activation='relu'))

# Eighth layer is a dense layer with 10 units for 10 classes in the output layer
model.add(Dense(num_classes, activation='softmax'))

model

optim = tf.keras.optimizers.Adadelta(
    learning_rate=0.1, 
    rho=0.95, 
    decay=0.0)

optim

# https://keras.io/optimizers/ 
model.compile(loss=keras.losses.categorical_crossentropy,
              optimizer=optim,
              metrics=['accuracy'])

model

model_history = \
    model.fit(x_train, y_train,
          batch_size=batch_size,
          epochs=epochs,
          verbose=2,
          validation_data=(x_test, y_test))

model_history.history['loss']

model_history.history['accuracy']

model.save("fitted_task2_model.keras")  # Save the model in HDF5 format

# Save the history
import json
with open("fitted_task2_model_history.json", "w") as f:
    json.dump(model_history.history, f)

#from keras.models import load_model

#model = load_model("fitted_baseline_model.keras")  # Load the saved model

# Load the history
#import json
#with open("fitted_baseline_model_history.json", "r") as f:
#    model_history = json.load(f)

model.summary()

loss_train, accuracy_train = model.evaluate(x_train, y_train)

print('Print the training loss and the accuracy of the model on the dataset')
print('Train Loss: {0:0.4f} Train Accuracy: {1:0.4f}'.format(loss_train, accuracy_train))

loss_test, accuracy_test = model.evaluate(x_test, y_test)

print('Print the testing loss and the accuracy of the model on the dataset')
print('Test Loss: {0:0.4f} Test Accuracy: {1:0.4f}'.format(loss_test, accuracy_test))

#score = model.evaluate(x_test, y_test, verbose=0)
#print('Test loss:', score[0])
#print('Test accuracy:', score[1])

########################################
# Making predictions and evaluating the model
#######################################

# Predicting the Test set results
y_pred = model.predict(x_test)
y_pred[y_pred > 0.9] = 1

y_pred[y_pred > 0.9] = 1
y_pred[y_pred <= 0.9] = 0

y_pred.shape

y_pred[0]

# summarize the middle 5 cases in a table

for i in range(15,20):
    print('actual   ', end=' ')
    for j in range(10):
        print('{:.1f}'.format(y_test[i][j]), end=' ')
    print()
    print('predicted', end=' ')
    for j in range(10):
        print('{:.1f}'.format(y_pred[i][j]), end=' ')
    print()
    print()

y_pred[0]

y_test[0]

y_pred.shape

y_test.shape

# Convert one-hot encoded arrays to label indices
y_pred_labels = np.argmax(y_pred, axis=1)
y_test_labels = np.argmax(y_test, axis=1)

y_pred_labels[0:15]

y_test_labels[0:15]

# Making the Confusion Matrix
cm = confusion_matrix(y_test_labels, y_pred_labels)
print('Print the Confusion Matrix:')
print(cm)

print(classification_report(y_test_labels, y_pred_labels))

disp = ConfusionMatrixDisplay(confusion_matrix=cm, display_labels=np.arange(10))
disp.plot()
plt.savefig('confusion_matrix_baseline.png')
plt.show()

print('Plot the accuracy')
# Keras 2.2.4 recognizes 'acc' and 2.3.1 recognizes 'accuracy'
# use the command python -c 'import keras; print(keras.__version__)' on MAC or Linux to check Keras' version
plt.plot(model_history.history['accuracy'], label='Training Accuracy')
#plt.plot(model_history.history['val_accuracy'], label='Validation Accuracy')
plt.title('5x5 Kernel Model Accuracy')
plt.ylabel('Accuracy')
plt.xlabel('Epoch')
plt.yticks(np.arange(0.9, 1.2, step=0.01))
plt.xticks(np.arange(0, 14, step=1))
plt.ylim(0.9, 1)
plt.xlim(0, 12)
plt.grid()
plt.legend(loc='lower right')
plt.savefig('task2_model_accuracy.png')
plt.show()

print('Plot the loss')
plt.plot(model_history.history['loss'], label='Training Loss')
#plt.plot(model_history.history['val_loss'], label='Validation Loss')
plt.title('5x5 Kernel Model Loss')
plt.ylabel('Loss')
plt.xlabel('Epoch')
plt.yticks(np.arange(0, 0.40, step=0.025))
plt.xticks(np.arange(0, 14, step=1))
plt.ylim(0, 0.35)
plt.xlim(0, 12)
plt.grid()
plt.legend(loc='upper right')
plt.savefig('task2_model_loss.png')
plt.show()