def distances_sum(point, point_array, distance_method=('euclidean', 2)):
distances = float(0)
m = np.shape(point_array)[0]
point = np.asarray(point)[0]
for j in range(m):
if distance_method[0] == 'euclidean':
distances += np.power(
distance.euclidean_distance(point,
np.asarray(point_array[j])[0]),
distance_method[1])
elif distance_method[0] == 'manhattan':
distances += np.power(
distance.manhattan_distance(point,
np.asarray(point_array[j])[0]),
distance_method[1])
elif distance_method[0] == 'minkowski':
distances += np.power(
distance.minkowski_distance(point,
np.asarray(point_array[j])[0],
distance_method[1]),
distance_method[1])
else:
distances += np.power(
distance.euclidean_distance(point, point_array[j]),
distance_method[1])
'''or
distance_j = distance.minkowski_distance(point, centroid_j, distance_method[1])
'''
return distances
def find_distance_min(point, point_array, distance_method=('euclidean', 2)):
min_distance = float('inf')
min_centroid_index = -1
k = np.shape(point_array)[0]
point = np.asarray(point)[0]
for j in range(k):
centroid_j = point_array[j]
if distance_method[0] == 'euclidean':
distance_j = distance.euclidean_distance(point, centroid_j)
elif distance_method[0] == 'manhattan':
distance_j = distance.manhattan_distance(point, centroid_j)
elif distance_method[0] == 'minkowski':
distance_j = distance.minkowski_distance(point, centroid_j,
distance_method[1])
else:
distance_j = distance.euclidean_distance(point, centroid_j)
'''or
distance_j = distance.minkowski_distance(point, centroid_j, distance_method[1])
'''
if distance_j < min_distance:
min_distance = distance_j
min_centroid_index = j
return min_centroid_index, min_distance
def window(prot1, prot2, instance, w):
di = dt.euclidean_distance(instance, prot1)
dj = dt.euclidean_distance(instance, prot2)
if di != 0 and dj != 0:
mini = min(di / dj, dj / di)
else:
mini = 0
s = ((1 - w) / (1 + w))
return (mini > s)
def __qp(self, X, kernel, C):
n_samples = X.shape[0]
P = 2 * kernel
q = -kernel[range(n_samples), range(n_samples)].reshape(-1, 1)
G = np.vstack((-np.eye(n_samples), np.eye(n_samples)))
h = np.hstack((np.zeros(n_samples), np.full(n_samples, C)))
A = np.full((1, n_samples), 1.0)
b = np.ones(1)
res = cvxopt.solvers.qp(cvxopt.matrix(P), cvxopt.matrix(q), cvxopt.matrix(G), cvxopt.matrix(h), cvxopt.matrix(A), cvxopt.matrix(b))
alpha = np.array(res['x']).ravel()
support_items = np.flatnonzero(np.isclose(alpha, 0) == False)
self.__X_support = X[support_items]
self.__a_support = alpha[support_items]
free_items = np.flatnonzero(self.__a_support < C)
X_free = self.__X_support[free_items]
self.__center = self.__a_support.dot(self.__X_support)
self.__radius = np.mean(distance.euclidean_distance(self.__center, X_free))
def kNN(k, data, instance):
"""
Returns a list with the k instances in the data set closest to a given instance
"""
# Extract the real data
classification = data['data']
# convert into float to be able to calculate euclidean distance
for index, li in enumerate(classification):
aux_list = list(float(x) for x in li[:-1])
classification[index] = aux_list + [li[-1]]
# initialize a dict with all votes to 0
classification_dict = {val: 0 for val in data['attributes'][-1][-1]}
# add the distance to the instance in a new field
for i in classification:
i.append(euclidean_distance(instance, i[:-1]))
# Sort the classification by the last element
sorted_classification = sorted(classification, key=operator.itemgetter(-1))
# we get the value of the K elements with shortest distance
for x in sorted_classification[:k]:
classification_dict[x[-2]] += 1
# Generate final candidates
candidates_list = list()
maximum_value = -1
for key, val in classification_dict.items():
if val > maximum_value:
candidates_list = [key]
maximum_value = val
elif val == maximum_value:
candidates_list.append(key)
return candidates_list
def k_neighbors(self, unknown, dataset, k):
"""
generate the closest neighbors list
"""
distances = []
for title in dataset:
point = dataset[title]
distance_to_point = distance.euclidean_distance(point, unknown)
distances.append([distance_to_point, title])
distances.sort()
neighbors = distances[0:k]
return neighbors
def fit(self, X, y, learning_rate, epochs):
'''
Parameters
----------
X : shape (n_samples, n_features)
Training data
y : shape (n_samples,)
Target values
learning_rate : learning rate
epochs : The number of epochs
'''
n_samples, n_features = X.shape
classes = np.unique(y)
n_classes = len(classes)
self.__prototypes = np.zeros((n_classes, n_features))
self.__prototypes_labels = np.zeros(n_classes)
for i in range(n_classes):
index_prototype = np.random.choice(np.flatnonzero(y == classes[i]),
1)
self.__prototypes[i] = X[index_prototype]
self.__prototypes_labels[i] = y[index_prototype]
for _ in range(epochs):
index = np.random.choice(n_samples, 1)
distances = distance.euclidean_distance(X[index],
self.__prototypes)
nearest_index = np.argmin(distances)
if self.__prototypes_labels[nearest_index] == y[index]:
self.__prototypes[nearest_index] += learning_rate * (
X[index] - self.__prototypes[nearest_index]).ravel()
else:
self.__prototypes[nearest_index] -= learning_rate * (
X[index] - self.__prototypes[nearest_index]).ravel()
def dist(x, y):
pos_d = euclidean_distance(x[:, :2], y[:2])
pix_d = euclidean_distance(x[:, 2:], y[2:])
d = np.sqrt(pos_d**2 + 5 * pix_d**2)
return d
def test_known3():
u = np.array([0, 0])
v = np.array([-3, -4])
assert_almost_equal(euclidean_distance(u, v), 5)
def test_known1():
u = np.array([0])
v = np.array([3])
assert_almost_equal(euclidean_distance(u, v), 3)
def test_triangle():
u = np.random.random(3)
v = np.random.random(3)
w = np.random.random(3)
assert euclidean_distance(
u, w) <= euclidean_distance(u, v) + euclidean_distance(v, w)
def test_symmetry():
for i in range(10):
u = np.random.random(3)
v = np.random.random(3)
assert euclidean_distance(u, v) == euclidean_distance(v, u)
def test_when_not_zero():
for i in range(10):
u = np.random.random(3)
v = np.zeros(3)
assert euclidean_distance(u, v) != 0
def test_when_zero():
u = np.zeros(3)
v = np.zeros(3)
assert euclidean_distance(u, v) == 0
def test_non_negative():
for i in range(10):
u = np.random.normal(3)
v = np.random.normal(3)
assert euclidean_distance(u, v) >= 0
from distance import euclidean_distance
if __name__ == '__main__':
pairs_of_points = [((0, 0), (0, 1)), ((0, 0), (1.5, 0)), ((0, 3), (4, 0)),
((0, 0), (1, 1)), ((-1, 0), (1, 1)), ((0, 0), (1, 3))]
distances = []
print 'The distances are:'
for pair in pairs_of_points:
distances.append(euclidean_distance(pair[0], pair[1]))
print distances
print 'The distances in ascending order are:'
distances.sort()
print distances
print 'The miminal distance in the list is:'
print distances[0]
print 'The two largest distances in the list are:'
print distances[-2], distances[-1]
print 'The distances whose round value is even are:'
print filter(lambda distance: distance % 2 == 0,
[round(distance) for distance in distances])
print("Tweet 1: " + str(z_stemmerem[0]) + ", TFIDF: " + str(bag_of_words[0]))
print()
print("Tweet 2: " + str(z_stemmerem[1]) + ", TFIDF: " + str(bag_of_words[1]))
print()
print("Tweet 3: " + str(z_stemmerem[2]) + ", TFIDF: " + str(bag_of_words[2]))
print()
print("Tweet 4: " + str(z_stemmerem[3]) + ", TFIDF: " + str(bag_of_words[3]))
print()
print("DYSTANS JACCARDA")
print(distance.jaccard_distance(bag_of_words[0], bag_of_words[1]))
print()
print("DYSTANS EUKLIDESA")
print(distance.euclidean_distance(bag_of_words[0], bag_of_words[1]))
print()
print("DYSTANS COSINE")
print(distance.cosine_distance(bag_of_words[0], bag_of_words[1]))
print()
print(nowynowy)
print("Demnostracja tf_idf")
print(calculate_tf_idf(tweet=nowynowy[0], corpus=nowynowy))
# kmeans test
print('----------------KMEANS --------------')
tweet1 = ['poland', 'poland', 'poland', 'poland', 'good', 'small', 'time']
tweet2 = ['poland', 'poland', 'poland', 'bad', 'big', 'coronavirus']
tweet3 = ['poland', 'poland', 'poland', 'growth', 'small', 'time']
tweet4 = ['poland', 'poland', 'poland', 'bad', 'big', 'virus']