plt.show()
def output_image(arr, m, n, filename):
arr = [255 if i == 1 else 0 for i in arr]
a = np.array(arr)
a = a.reshape((m, n))
cv2.imwrite(filename, a)
if __name__ == "__main__":
#####################
# read files
colors = ["red", "blue"]
k = 2
circs_X = read_file("data.data")
# Q1 circs
#####################
# regular k-means
kmeans_model1 = KMeans(n_clusters=k,
random_state=np.random.randint(100)).fit(circs_X)
kmeans_labels1 = kmeans_model1.labels_
# show(circs_X, kmeans_labels1)
print(calc_loss(circs_X, kmeans_labels1))
# spectral clustering
# sigma1 = 0.01
# sigma1 = 0.1
# sigma1 = 1
# sigma1 = 5
return linear_kernel
@staticmethod
def gauss_kernel_generator(sigma2):
def gauss_kernel(x1, x2):
return np.exp(-np.linalg.norm(x1 - x2) ** 2 / (2. * sigma2))
return gauss_kernel
def distance(hyperplane, point):
return np.inner(hyperplane.flatten(), point) / np.linalg.norm(hyperplane[1:-1])
if __name__ == "__main__":
# read data
train_X, train_Y = read_file("train.data")
validation_X, validation_Y = read_file("validation.data")
test_X, test_Y = read_file("test.data")
train_len = train_X.shape[0]
validation_len = validation_X.shape[0]
test_len = test_X.shape[0]
#
c_arr = [1, 10, 100, 1000, 1e4, 1e5, 1e6, 1e7, 1e8]
sigma2_arr = [0.1, 1, 10, 100, 1000]
#
s = dict()
for c in c_arr:
for sigma2 in sigma2_arr:
# s[c, sigma2] = SVM("spam_train.data", c=c, sigma2=sigma2)
print("c: ", c, " sigma2: ", sigma2)
@staticmethod
def predict(dt, X):
predicts = []
for i in range(X.shape[0]):
x = X[i]
predicts += [dt.predict_single(x)]
return predicts
@staticmethod
def calc_accu(t, test_X, test_Y):
pred = DecisionTree.predict(t, test_X)
accu = np.sum(test_Y == pred) / test_Y.shape[0]
return accu
if __name__ == "__main__":
feature_names = np.array([
'cap-shape', 'cap-surface', 'cap-color', 'bruises?', 'odor',
'gill-attachment', 'gill-spacing', 'gill-size', 'gill-color',
'stalk-shape', 'stalk-root', 'stalk-surface-above-ring',
'stalk-surface-below-ring', 'stalk-color-above-ring',
'stalk-color-below-ring', 'veil-type', 'veil-color', 'ring-number',
'ring-type', 'spore-print-color', 'population', 'habitat'
])
train_X, train_Y = read_file("train.data")
test_X, test_Y = read_file("test.data")
t = DecisionTree(train_X, train_Y, feature_names)
accu = DecisionTree.calc_accu(t, test_X, test_Y)
# t subtree
tsub = t.tsub
G.add_edge(t.name, tsub.name)
e = G.get_edge(t.name, tsub.name)
e.attr['label'] = "t"
draw_tree_helper(G, tsub)
# f subtree
fsub = t.fsub
G.add_edge(t.name, fsub.name)
e = G.get_edge(t.name, fsub.name)
e.attr['label'] = "f"
draw_tree_helper(G, fsub)
if __name__ == "__main__":
heart_train_X, heart_train_Y = read_file("train.data")
heart_test_X, heart_test_Y = read_file("test.data")
# 1 (a)
epsilons1, alphas1, predictors1 = adaboost(heart_train_X, heart_train_Y, 5)
ada_pred_Y = adaboost_pred(alphas1, predictors1, heart_train_X)
accu = np.sum(ada_pred_Y == heart_train_Y) / heart_train_Y.size
# draw tree
for i in range(len(predictors1)):
t = predictors1[i]
draw_tree(t, str(i))
# 1 (b)
epsilons2, alphas2, predictors2 = adaboost(heart_train_X, heart_train_Y,
for i in range(X.shape[0]):
xi = X[i]
labeli = labels[i]
s += np.sum((xi - mus[labeli])**2)
return s
def func_uniform_draw(left, right):
def draw(m, n):
return np.random.uniform(left, right, (m, n))
return draw
if __name__ == "__main__":
train_X, train_Y = read_file("train.data")
func_draw = func_uniform_draw(-3, 3)
ks = [12, 18, 24, 36, 42]
# k-means
k1, g1 = dict(), dict()
for k in ks:
print(k)
k1[k], g1[k] = [], []
for i in range(20):
init_centres = func_draw(k, train_X.shape[1])
k1[k] += [KMeans(n_clusters=k, init=init_centres).fit(train_X)]
g1[k] += [GMM(train_X, k, mu=init_centres)]
# k-means++, GMM++
k2, g2 = dict(), dict()