OI#2

oi.py

import matplotlib.pyplot as plt
import numpy as np
import h5py
import os
from sklearn.neural_network import MLPClassifier
from sklearn.svm import SVC


cwd = os.getcwd()  # current working directory
path = os.path.join(cwd, 'data')


def load_dataset():
    file_name = os.path.join(path, 'train_catvnoncat.h5')
    train_dataset = h5py.File(file_name, "r")
    x_train = np.array(train_dataset["train_set_x"][:])  # your train set features
    y_train = np.array(train_dataset["train_set_y"][:])  # your train set labels

    file_name = os.path.join(path, 'test_catvnoncat.h5')
    test_dataset = h5py.File(file_name, "r")
    x_test = np.array(test_dataset["test_set_x"][:])  # your test set features
    y_test = np.array(test_dataset["test_set_y"][:])  # your test set labels

    classes = ['non-cat', 'cat']

    y_train = y_train.reshape(-1, 1)
    y_test = y_test.reshape(-1, 1)

    return x_train, y_train, x_test, y_test, classes


x_train, y_train, x_test, y_test, classes = load_dataset()

index = 27
plt.figure()
plt.imshow(x_train[index])
plt.show()
print("y = " + str(y_train[index, :]) + ", it's a '" + classes[np.squeeze(y_train[index, :])] + "' picture.")

# Normalize data
m_train, num_px, _, _ = x_train.shape
m_test = x_test.shape[0]
x_train_flatten = x_train.reshape(m_train, -1)
x_test_flatten = x_test.reshape(m_test, -1)

x_train_scaled = x_train_flatten / 255.
x_test_scaled = x_test_flatten / 255.

y_train = np.squeeze(y_train)
y_test = np.squeeze(y_test)
print('Y_train.shape =', y_train.shape)
print('Y_test.shape =', y_test.shape)
print('X_train_scaled.shape =', x_train_scaled.shape)

for activation in ['logistic', 'relu']:
    # 1-layer NN classifier
    print('- NN 1,', activation)
    clf = MLPClassifier(
        hidden_layer_sizes=100,
        solver='sgd',
        activation=activation,
        max_iter=300,
        alpha=0.0001
    ).fit(x_train_scaled, y_train)

    print("train accuracy= {:.3%}".format(clf.score(x_train_scaled, y_train)))
    print("test accuracy= {:.3%}".format(clf.score(x_test_scaled, y_test)))

    # 2-layers MLP classifier
    print('- NN 2', activation)
    clf = MLPClassifier(
        hidden_layer_sizes=(3, 3),
        solver='sgd',
        activation='logistic',
        max_iter=300,
        alpha=0.0001
    ).fit(x_train_scaled, y_train)

    print("train accuracy = {:.3%}".format(clf.score(x_train_scaled, y_train)))
    print("test accuracy = {:.3%}".format(clf.score(x_test_scaled, y_test)))

    # 3-layers MLP classifier
    print('- NN 3', activation)
    clf = MLPClassifier(
        hidden_layer_sizes=(20, 7, 10),
        solver='sgd',
        activation='logistic',
        max_iter=300,
        alpha=0.0001
    ).fit(x_train_scaled, y_train)

    print("train accuracy = {:.3%}".format(clf.score(x_train_scaled, y_train)))
    print("test accuracy = {:.3%}".format(clf.score(x_test_scaled, y_test)))

# Compare with SVM
print('SVM')
clf = SVC(C=10, kernel='rbf', gamma=0.001).fit(x_train_scaled, y_train)
print("train accuracy = {:.3%}".format(clf.score(x_train_scaled, y_train)))
print("test accuracy = {:.3%}".format(clf.score(x_test_scaled, y_test)))