train_y_file = "./data/train_label.csv"
train_y = np.genfromtxt(train_y_file, delimiter=',')
print(train_y.shape)
test_x_file = "./data/test_data.csv"
test_x = np.genfromtxt(test_x_file, delimiter=',')
print(test_x.shape)
test_y_file = "./data/test_label.csv"
test_y = np.genfromtxt(test_y_file)
print(test_y.shape)
#addition PCA here
pca = PCA(166, 50)
train_x = pca.getPC(train_x)
test_x = pca.projectData(
test_x) # standardization will be performed inside the function
print(train_x.shape)
print(test_x.shape)
knn = Knn()
print type(train_x)
new_train_x, new_train_y = knn.shuffle(train_x, train_y)
new_train_x = np.array(new_train_x)
new_train_y = np.array(new_train_y)
k_values = [2 * l + 1 for l in range(0, 9)]
k_value_acc = {i: 0.0 for i in k_values}
for k in k_values:
sum_accuracy = 0
folds = 5
for i in range(1, folds + 1):
train_x1, train_y1, test_x1, test_y1 = knn.get_kth_fold(
new_train_x, new_train_y, folds, i)
p1 = mu
p2 = mu + 1.5 * S[0] * U[:, 0].T
data = np.r_[p1, p2].reshape([-1, 2])
ax.plot(data[:, 0], data[:, 1], '-k', linewidth=2)
p2 = mu + 1.5 * S[1] * U[:, 1].T
data = np.r_[p1, p2].reshape([-1, 2])
ax.plot(data[:, 0], data[:, 1], '-k', linewidth=2)
# =================== Dimension Reduction ===================
ax = fig.add_subplot(1, 2, 2)
ax.plot(X_norm[:, 0], X_norm[:, 1], 'bo')
K = 1
Z = pca.projectData(X_norm, U, K)
X_rec = pca.recoverData(Z, U, K)
ax.plot(X_rec[:, 0], X_rec[:, 1], 'ro')
for i in range(m):
data = np.r_[X_norm[i, :], X_rec[i, :]].reshape([-1, 2])
ax.plot(data[:, 0], data[:, 1], '--k', linewidth=2)
# =============== Loading and Visualizing Face Data =============
def displayData(X):
m, n = X.shape
example_width = np.int(np.sqrt(n))
example_height = np.int(n / example_width)
