def __init__(self, number_of_gaussian):
self.ng = number_of_gaussian
self.gmm = GMM(number_of_gaussian)
self.comparator_with_impostor = ComparatorWithImpostor(
number_of_gaussian)
self.comparator = Comparator(number_of_gaussian)
self.cross_validator = CrossValidation(0.3)
class trainingClass:
def __init__(self, size, output_wavefile):
self.Training_info = GMM(size, output_wavefile);
def Training_feature_Mean(self):
Mean_training = self.Training_info.GMM_Model_Mean()
return Mean_training
def Training_feature_Weight(self):
Weight_training = self.Training_info.GMM_Model_Weight()
return Weight_training
def Training_feature_Covar(self):
Covar_training = self.Training_info.GMM_Model_Covar()
return Covar_training
def adjustFeatures(self,name,mainF):
self.Training_info.adjustFeatures(name,mainF);
def cute_test(length):
p_array = []
counter = 0
for i in range(0, 1000):
#cause, effect = generate_continue_data(200, random.randint(1,3)) # random.randint(1,5)
cause = GMM(5, 200)
effect = GMM(8, 200)
#effect, test2 = generate_continue_data(200, 3) # random.randint(1,5)
cause = change_to_zero_one(cause)
effect = change_to_zero_one(effect)
cause2effect = bernoulli2(effect, length) - cbernoulli2(
effect, cause, length)
effect2cause = bernoulli2(cause, length) - cbernoulli2(
cause, effect, length)
#print 'cause' + ' -> ' + 'effect' + ':' + str(cause2effect)
#print 'effect' + ' -> ' + 'cause' + ':' + str(effect2cause)
p = math.pow(2, -(cause2effect - effect2cause))
p_array.append(p)
if cause2effect > effect2cause:
counter += 1
print
print counter
print bh_procedure(p_array, 0.05)
return counter / 100.0
def main():
pref_path = os.getcwd() + "/classification_data_HWK2/EMGaussian"
train_data = np.loadtxt(open(pref_path + ".data", "rb"), delimiter=" ")
test_data = np.loadtxt(open(pref_path + ".test", "rb"), delimiter=" ")
Xtrain = train_data[:, :2]
Xtest = test_data[:, :2]
models = {"GMM": GMM(isotropic=False), "HMM": HMM()}
K = 4 #number of clusters
for name in ["GMM", "HMM"]:
print(name)
model = models[name]
model.fit(Xtrain, K, eps=pow(10, -2))
# visualize clusters and frontiers
model.plot_clusters(Xtrain, "figs/" + name + " on train", save=True)
model.plot_clusters(Xtest, "figs/" + name + " on test", save=True)
print("")
lik = model.compute_log_likelihood(Xtrain)
print("mean log-likelihood on training set : ", lik / Xtrain.shape[0])
lik = model.compute_log_likelihood(Xtest)
print("mean log-likelihood on test set : ", lik / Xtest.shape[0])
print("\n------------------------\n")
def main():
trainD, devD, testD = init()
allD = Data(trainD, devD)
if sys.argv[1] == "display":
display(allD)
exit(0)
if sys.argv[1] == "train":
# local settings
x = trainD.nx()
y = trainD.ny()
xx = devD.nx()
else:
# submit settings
x = allD.nx()
y = allD.ny()
xx = testD.nx()
gmm1 = GMM(x[y == 1], round=500, K=4)
gmm2 = GMM(x[y == 2], round=500, K=4)
print("GMM1.dist: ", gmm1.pi)
print("GMM2.dist: ", gmm2.pi)
r1 = gmm1.predict(xx) * np.sum(y == 2)
r2 = gmm2.predict(xx) * np.sum(y == 1)
result = 1 + (r1 < r2) * 1
if sys.argv[1] == "train":
# local settings
print("accuracy: ", sum(result == devD.ny()) / devD.ny().shape[0])
else:
# submit settings
testD.y = list(result)
testD.output()
def test_cluster_num():
for i in [2,3,4,5,6]:
Data = np.load("./Data/cluster_"+str(i)+".npy")
gmm = GMM(data = Data)
gmm.aic()
gmm.bic()
vbem = VBEM(data = Data)
vbem.train()
vbem.show()
def test_sample_sizes():
for i in [10,20,50,100,200,300]:
Data = np.load("./Data/data_"+str(i)+".npy")
gmm = GMM(data = Data)
gmm.aic()
gmm.bic()
vbem = VBEM(data = Data)
vbem.train()
vbem.show()
def test_dimension():
for i in [3,4,5,6,7]:
Data = np.load("./Data/dimension_"+str(i)+".npy")
gmm = GMM(data = Data)
gmm.aic()
gmm.bic()
vbem = VBEM(data = Data)
vbem.train()
vbem.show()
def main():
pref_path = os.getcwd() + "/classification_data_HWK2/EMGaussian"
train_data = np.loadtxt(open(pref_path + ".data", "rb"), delimiter=" ")
test_data = np.loadtxt(open(pref_path + ".test", "rb"), delimiter=" ")
Xtrain = train_data[:, :2]
Xtest = test_data[:, :2]
models = {
"Kmeans": Kmeans(),
"GMM_general": GMM(isotropic=False),
"GMM_isotropic": GMM(isotropic=True)
}
K = 4 #number of clusters
for name in ["Kmeans", "GMM_isotropic", "GMM_general"]:
print(name)
model = models[name]
model.fit(Xtrain, 4)
# visualize clusters and frontiers
model.plot_clusters(Xtrain, name + " on train", save=False)
model.plot_clusters(Xtest, name + " on test", save=False)
if name in ["GMM_general", "GMM_isotropic"]:
lik = model.compute_log_likelihood(Xtrain)
print("mean log-likelihood on training set : ",
lik / Xtrain.shape[0])
lik = model.compute_log_likelihood(Xtest)
print("mean log-likelihood on test set : ", lik / Xtest.shape[0])
print("")
def main():
# declare variables
K_value = 3
max_epoch = 1000
repeat_num = 8
repeat_num_gmm = 1
name = 'GMM_dataset.txt'
# get the training data
training_data = data_loading(name)
if (sys.argv[1] == 'kmean'):
# run the K-mean program
program_name = './a.out'
parameter_line = ' ' + 'training_kmeans ' + str(K_value) + ' ' + str(
repeat_num) + ' 0'
print('Running K-mean')
os.system(program_name + parameter_line)
print('The Program is done.')
# read minimum SSE Position
min_sse_pos = np.loadtxt('./min_sse_pos.csv',
delimiter=',',
skiprows=0)
# read clusters from K-mean algorithm
kmean_name = './all_cluster_center' + str(int(min_sse_pos)) + '.csv'
kmean_clusters = np.loadtxt(kmean_name, delimiter=',', skiprows=0)
for draw_ind in range(int(kmean_clusters.shape[0] / K_value)):
# the last index: (int(kmean_clusters.shape[0] / K_value) - 1)
start_index = draw_ind * K_value
iter_kmean_clusters = kmean_clusters[start_index:start_index +
K_value, :]
# label assignment
out_label, cov_list = label_assignment(iter_kmean_clusters,
training_data)
# save the figures
plt.rcParams.update({'figure.max_open_warning': 0})
fig, ax = plt.subplots()
ax.scatter(training_data[:, 0],
training_data[:, 1],
c=out_label,
alpha=0.5)
ax.scatter(iter_kmean_clusters[:, 0],
iter_kmean_clusters[:, 1],
c='b',
s=100,
alpha=0.5)
plt.xlabel('Feature: x1')
plt.ylabel('Feature: x2')
plt.title('K-mean Clustering')
fig.savefig('./kmean_result/iter' + str(draw_ind) + '.png')
#fig.clf()
elif (sys.argv[1] == 'gmm'):
# read minimum SSE Position
min_sse_pos = np.loadtxt('./min_sse_pos.csv',
delimiter=',',
skiprows=0)
# read clusters from K-mean algorithm
kmean_name = './all_cluster_center' + str(int(min_sse_pos)) + '.csv'
kmean_clusters = np.loadtxt(kmean_name, delimiter=',', skiprows=0)
start_index = (int(kmean_clusters.shape[0] / K_value) - 1) * K_value
iter_kmean_clusters = kmean_clusters[start_index:start_index +
K_value, :]
out_label, cov_list = label_assignment(iter_kmean_clusters,
training_data)
# call GMM class
gmm = GMM(K_value, repeat_num_gmm, max_epoch, training_data)
all_likelihood, parameters = gmm.model_training(
iter_kmean_clusters, out_label, cov_list)
true_mu, true_covariance = gmm.fit_true_model(training_data)
# find the mu, covariance, and prior
best_likelihood, all_mu, all_cov, all_prior = find_the_best(
all_likelihood, parameters)
# prediction phase
prediction = gmm.model_predict(all_mu[-1], all_cov[-1], all_prior[-1],
training_data)
labels = label_GMM(prediction)
# drawing gaussian functions
for ind in range(len(all_mu)):
out_para = drawing_Gaussian(all_mu[ind], all_cov[ind],
training_data, all_mu[-1], 1, ind)
if ind == (len(all_mu) - 1):
# print out parameters of the Gaussian function
for ind_2 in range(K_value):
print('Cluster:', ind_2)
print('Mu:')
print(out_para[str(ind_2)][0])
print('Covariance:')
print(out_para[str(ind_2)][1])
print('=====================')
# drawing true gaussian functions
if K_value == 3:
out_param_true = drawing_Gaussian(true_mu, true_covariance,
training_data, true_mu, 2, None)
# print out parameters of the Gaussian function
for ind_3 in range(K_value):
print('Cluster:', ind_3)
print('Actual Mu:')
print(out_param_true[str(ind_3)][0])
print('Actual Covariance:')
print(out_param_true[str(ind_3)][1])
print('=====================')
# drawing log-likelihood values
drawing_Log_likelihood(best_likelihood)
elif (sys.argv[1] == 'saving'):
# save the data as a csv file
save_data_csv_file(training_data)
else:
print('Error Input. Please re-choose the task!!')
mask = np.zeros_like(img)
mask[:, :, 0] = probsReshaped
mask[:, :, 1] = probsReshaped
mask[:, :, 2] = probsReshaped
return mask
if __name__ == '__main__':
# Loading models
path = 'models/'
files = os.listdir(path)
paths = ['Green_Resized/', 'Orange_Resized/', 'Yellow_Resized/']
gmms = [GMM(nClusters=3), GMM(nClusters=3), GMM(nClusters=3)] # TODO
gParams, gMixture = None, None
oParams, oMixture = None, None
yParams, yMixture = None, None
for file in files:
filename = os.path.join(path, file)
pickle_in = open(filename, "rb")
model = pickle.load(pickle_in)
for key, info in model.items():
if key == paths[0]:
gParams = info[0]
gMixture = info[1]
@brief TBD
@license This project is released under the BSD-3-Clause license.
'''
import numpy as np
import cv2
from scipy.stats import multivariate_normal
from GMM import GMM
#image = cv2.imread('training_set/yellowTrainingSet.png')
image = cv2.imread('test_set/buoys.png')
height = image.shape[0]
width = image.shape[1]
yellowGMM = GMM()
yellowGMM.load('yellowGMM.npz')
orangeGMM = GMM()
orangeGMM.load('orangeGMM.npz')
greenGMM = GMM()
greenGMM.load('greenGMM.npz')
'''
yellowErrors = np.zeros((height, width))
orangeErrors = np.zeros((height, width))
greenErrors = np.zeros((height, width))
for i in range(height):
for j in range(width):
yellowErros[i] = yellowGMM.getLogLikelihoodError(image[i,j])
# connectivity matrix for structured Ward
connectivity = kneighbors_graph(X,
n_neighbors=params['n_neighbors'],
include_self=False)
# make connectivity symmetric
connectivity = 0.5 * (connectivity + connectivity.T)
# ============
# 初始化所有聚类算法
# ============
# 自编的K-Means、GMM算法
my_kmeans0 = K_Means(n_clusters=params['n_clusters'], fit_method=0)
my_kmeans1 = K_Means(n_clusters=params['n_clusters'], fit_method=1)
my_kmeans2 = K_Means(n_clusters=params['n_clusters'], fit_method=2)
my_kmeans3 = K_Means(n_clusters=params['n_clusters'], fit_method=3)
my_gmm = GMM(n_clusters=params['n_clusters'], dim=X.shape[1])
my_spectral_knn_reciprocal_normalized = SpectralClustering(
n_clusters=params['n_clusters'], nnk=50)
my_spectral_radius_reciprocal_normalized = SpectralClustering(
n_clusters=params['n_clusters'], use_radius_nn=True, nnradius=1)
my_spectral_knn_gauss05_normalized = SpectralClustering(
n_clusters=params['n_clusters'], nnk=50, use_gauss_dist=True)
my_spectral_knn_gauss005_normalized = SpectralClustering(
n_clusters=params['n_clusters'],
nnk=50,
use_gauss_dist=True,
gauss_sigma=5e-2)
my_spectral_knn_reciprocal_unnormalized = SpectralClustering(
n_clusters=params['n_clusters'], nnk=50, normalized=False)
# sklearn中自带的算法
import numpy as np
import matplotlib.pyplot as plt
from GMM import GMM
if __name__ == '__main__':
group_a = np.random.normal(loc=(20.00, 14.00), scale=(4.0, 4.0), size=(1000, 2))
group_b = np.random.normal(loc=(15.00, 8.00), scale=(2.0, 2.0), size=(1000, 2))
group_c = np.random.normal(loc=(30.00, 40.00), scale=(2.0, 2.0), size=(1000, 2))
group_d = np.random.normal(loc=(25.00, 32.00), scale=(7.0, 7.0), size=(1000, 2))
group_e = np.random.normal(loc=(10.00, 32.00), scale=(7.0, 7.0), size=(1000, 2))
DATA = np.concatenate((group_a, group_b, group_c, group_d, group_e))
S = GMM(5, DATA, 1e-3)
S.fit()
S.print_status()
testdata = np.random.rand(10000, 2)*50
labels = S.Classify(testdata)
plt.scatter(testdata[:, 0], testdata[:, 1], c=list(map(lambda i : {0:'b',1:'g',2:'r',3:'y',4:'k'}[i], labels)))
plt.show()
"""
chapter9 EM algorithm for Gaussian Misture Model
Using iris dataset for clustering
"""
import DatasetUtil as DS
from HTMLTable import HTMLTable
import re
from GMM import GMM
if __name__ == "__main__":
print("\t============ Chap9 EM for GMM ============")
ds = DS.DATAUtil()
x_train, y_train = ds.load(True, r".\dataset.dat")
model = GMM()
model.train(x_train)
y_pred = model.predict(x_train)
y_train = ds.y_int2str(y_train)
table = HTMLTable(caption='Iris Data Cluster')
table.append_header_rows((
('No.', 'A1', 'A2', 'A3', 'A4', 'Classification', ''),
('', '', '', '', '', 'Label-C', 'Predict-C'),
))
table[0][0].attr.rowspan = 2
table[0][1].attr.rowspan = 2
table[0][2].attr.rowspan = 2
table[0][3].attr.rowspan = 2
table[0][4].attr.rowspan = 2
from GMM import GMM
from sklearn.datasets import make_blobs
import matplotlib.pyplot as plt
X, y = make_blobs(n_samples=1000, centers=4, n_features=2)
gmm_cls = GMM(initializer='uniform', cov_type='diag')
gmm_cls.fit(X, 4)
colors = []
for l in gmm_cls.kmeans_cls_.predict(X):
if l == 0:
colors.append('red')
if l == 1:
colors.append('green')
if l == 2:
colors.append('orange')
if l == 3:
colors.append('blue')
plt.scatter(X[:, 0], X[:, 1], c=colors, alpha=0.1)
plt.scatter(gmm_cls.means_[:, 0], gmm_cls.means_[:, 1], c='k')
plt.show()
plt.scatter(X[:, 0], X[:, 1], c=colors, alpha=0.1)
plt.scatter(gmm_cls.kmeans_cls_.means_[:, 0],
gmm_cls.kmeans_cls_.means_[:, 1],
c='k')
plt.show()
def fit3(self, raw_colors):
self._X = [raw_colors[i] for i in Classifier.FACELET_ORDER]
np.save('X', self._X)
X = np.load('X.npy').astype('uint8')
for i in range(len(X)):
X[i][0], X[i][2] = X[i][2], X[i][0]
X_hsv = X[np.newaxis, ...]
hsv = cv2.cvtColor(X_hsv, cv2.COLOR_RGB2HSV)[0]
centroid_id = list(range(4, 54, 9))
feats = hsv[:, ::2]
ax = pl.subplot(111)
for i, c in enumerate(feats):
if i % 9 == 4:
# print(type(raw_colors[i]))
pl.scatter(*feats[i], color=X[i] / 255.0, marker='x')
else:
pl.scatter(*feats[i], color=X[i] / 255.0, marker='o')
pl.show()
alphas = np.empty((6, 1))
means = np.empty((6, feats.shape[1]))
covs = np.empty((6, feats.shape[1], feats.shape[1]))
for i in range(6):
mean = feats[9 * i + 4]
cov = np.array([[0.5 * 10, 0.0], [0.0, 2.5 * 10]])
alphas[i] = 1 / 6.0
means[i] = mean
covs[i] = cov
print('alpha:', alphas[i])
print('mean:', mean)
print('cov:', cov)
self._gmm = GMM(6, feats, mu=means, sigma=covs, alpha=alphas)
self._gmm.execute()
print(self._gmm.alpha)
print(self._gmm.mu)
print(self._gmm.sigma)
self._ans = []
for xi in feats:
probas = np.array([
self._gmm.Normal(xi, self._gmm.mu[k], self._gmm.sigma[k],
len(xi)) for k in range(6)
])
self._ans.append(probas.argmax())
print(np.array(self._ans).reshape(6, 3, 3))
def eigsorted(cov):
vals, vecs = np.linalg.eigh(cov)
order = vals.argsort()[::-1]
return vals[order], vecs[:, order]
nstd = 2
ax = pl.subplot(111)
for i, c in enumerate(self._ans):
if i % 9 == 4:
pl.scatter(*feats[i],
color=X[centroid_id[c]] / 255.0,
marker='x')
else:
pl.scatter(*feats[i],
color=X[centroid_id[c]] / 255.0,
marker='o')
for k in range(6):
cov = self._gmm.sigma[k]
vals, vecs = eigsorted(cov)
theta = np.degrees(np.arctan2(*vecs[:, 0][::-1]))
w, h = 2 * nstd * np.sqrt(vals)
ell = mpl.patches.Ellipse(xy=self._gmm.mu[k],
width=w,
height=h,
angle=theta,
color=X[centroid_id[k]] / 255.0,
alpha=0.5)
ell.set_facecolor(X[centroid_id[k]] / 255.0)
ax.add_artist(ell)
for k in range(6):
cov = covs[k]
vals, vecs = eigsorted(cov)
theta = np.degrees(np.arctan2(*vecs[:, 0][::-1]))
w, h = 2 * nstd * np.sqrt(vals)
ell = mpl.patches.Ellipse(xy=self._gmm.mu[k],
width=w,
height=h,
angle=theta,
color=X[centroid_id[k]] / 255.0,
alpha=0.5)
# ell.set_facecolor(X[centroid_id[k]]/255.0)
ax.add_artist(ell)
pl.show()
class Classifier(object):
"""docstring for Classifier."""
FACELET_ORDER = [
42,
39,
36,
43,
40,
37,
44,
41,
38, # U
9,
10,
11,
12,
13,
14,
15,
16,
17, # R
8,
7,
6,
5,
4,
3,
2,
1,
0, # F
45,
46,
47,
48,
49,
50,
51,
52,
53, # D
27,
28,
29,
30,
31,
32,
33,
34,
35, # L
26,
25,
24,
23,
22,
21,
20,
19,
18, # B
]
def __init__(self, face_order='URFDLB'):
self._fig = pl.figure()
self._axis = self._fig.add_subplot(111, projection='3d')
self._face_order = face_order
def fit(self, raw_colors):
self._X = [raw_colors[i] for i in Classifier.FACELET_ORDER]
np.save('X', self._X)
for label, color in enumerate(self._X):
color[0], color[2] = color[2], color[0]
self._axis.scatter(*color, color=color / 384)
pl.show()
self._X_pca = PCA(n_components=2).fit_transform(self._X)
print(self._X_pca)
np.save('pca', self._X_pca)
for i, point in enumerate(self._X_pca):
pl.scatter(*point, color=self._X[i] / 384)
pl.show()
centroids = np.array([self._X_pca[i] for i in range(4, 54, 9)])
for i, point in enumerate(centroids):
pl.scatter(*point, color=self._X[9 * i + 4] / 384)
pl.show()
print(centroids)
self._model = KMeans(n_clusters=6, init=centroids)
# self._model = GaussianMixture(n_components=6, means_init=centroids)
self._model.fit(self._X_pca)
def fit2(self, raw_colors):
from sklearn.mixture import GaussianMixture
self._X = [raw_colors[i] for i in Classifier.FACELET_ORDER]
for label, color in enumerate(self._X):
color[0], color[2] = color[2], color[0]
self._axis.scatter(*color, color=color / 384)
pl.show()
centroids = np.array([self._X[i] for i in range(4, 54, 9)])
print(centroids)
self._model = GaussianMixture(n_components=6, means_init=centroids)
self._model.fit(self._X)
self._X_pca = self._X
def fit3(self, raw_colors):
self._X = [raw_colors[i] for i in Classifier.FACELET_ORDER]
np.save('X', self._X)
X = np.load('X.npy').astype('uint8')
for i in range(len(X)):
X[i][0], X[i][2] = X[i][2], X[i][0]
X_hsv = X[np.newaxis, ...]
hsv = cv2.cvtColor(X_hsv, cv2.COLOR_RGB2HSV)[0]
centroid_id = list(range(4, 54, 9))
feats = hsv[:, ::2]
ax = pl.subplot(111)
for i, c in enumerate(feats):
if i % 9 == 4:
# print(type(raw_colors[i]))
pl.scatter(*feats[i], color=X[i] / 255.0, marker='x')
else:
pl.scatter(*feats[i], color=X[i] / 255.0, marker='o')
pl.show()
alphas = np.empty((6, 1))
means = np.empty((6, feats.shape[1]))
covs = np.empty((6, feats.shape[1], feats.shape[1]))
for i in range(6):
mean = feats[9 * i + 4]
cov = np.array([[0.5 * 10, 0.0], [0.0, 2.5 * 10]])
alphas[i] = 1 / 6.0
means[i] = mean
covs[i] = cov
print('alpha:', alphas[i])
print('mean:', mean)
print('cov:', cov)
self._gmm = GMM(6, feats, mu=means, sigma=covs, alpha=alphas)
self._gmm.execute()
print(self._gmm.alpha)
print(self._gmm.mu)
print(self._gmm.sigma)
self._ans = []
for xi in feats:
probas = np.array([
self._gmm.Normal(xi, self._gmm.mu[k], self._gmm.sigma[k],
len(xi)) for k in range(6)
])
self._ans.append(probas.argmax())
print(np.array(self._ans).reshape(6, 3, 3))
def eigsorted(cov):
vals, vecs = np.linalg.eigh(cov)
order = vals.argsort()[::-1]
return vals[order], vecs[:, order]
nstd = 2
ax = pl.subplot(111)
for i, c in enumerate(self._ans):
if i % 9 == 4:
pl.scatter(*feats[i],
color=X[centroid_id[c]] / 255.0,
marker='x')
else:
pl.scatter(*feats[i],
color=X[centroid_id[c]] / 255.0,
marker='o')
for k in range(6):
cov = self._gmm.sigma[k]
vals, vecs = eigsorted(cov)
theta = np.degrees(np.arctan2(*vecs[:, 0][::-1]))
w, h = 2 * nstd * np.sqrt(vals)
ell = mpl.patches.Ellipse(xy=self._gmm.mu[k],
width=w,
height=h,
angle=theta,
color=X[centroid_id[k]] / 255.0,
alpha=0.5)
ell.set_facecolor(X[centroid_id[k]] / 255.0)
ax.add_artist(ell)
for k in range(6):
cov = covs[k]
vals, vecs = eigsorted(cov)
theta = np.degrees(np.arctan2(*vecs[:, 0][::-1]))
w, h = 2 * nstd * np.sqrt(vals)
ell = mpl.patches.Ellipse(xy=self._gmm.mu[k],
width=w,
height=h,
angle=theta,
color=X[centroid_id[k]] / 255.0,
alpha=0.5)
# ell.set_facecolor(X[centroid_id[k]]/255.0)
ax.add_artist(ell)
pl.show()
def get_state(self):
# if self._X is None:
# raise Error('Se debe hacer fit primero.')
# pred2 = self._model.predict(self._X_pca)
# state = ''.join([self._face_order[i] for i in pred2])
# print(state)
# pred2 = pred2.reshape(-1, 3, 3)
# print(pred2)
# return state
FACES = 'URFDLB'
state = ''.join([FACES[i] for i in np.array(self._ans).reshape(-1)])
print(state)
return state
class BuoyDetector:
yellowGMM = GMM()
orangeGMM = GMM()
greenGMM = GMM()
def __init__(self, yellowGMMParams, orangeGMMParams, greenGMMParams):
self.yellowGMM.load(yellowGMMParams)
self.orangeGMM.load(orangeGMMParams)
self.greenGMM.load(greenGMMParams)
def detectBuoys(self, frame):
image = frame.copy()
height = image.shape[0]
width = image.shape[1]
#print('BP1')
yellowMask = np.zeros((height, width)).astype('uint8')
orangeMask = np.zeros((height, width)).astype('uint8')
greenMask = np.zeros((height, width)).astype('uint8')
#print('BP2')
yellowErrors = self.yellowGMM.getLogLikelihoodError(np.reshape(image, (height*width, 3)))
orangeErrors = self.orangeGMM.getLogLikelihoodError(np.reshape(image, (height*width, 3)))
greenErrors = self.greenGMM.getLogLikelihoodError(np.reshape(image, (height*width, 3)))
yellowErrors = np.reshape(yellowErrors, (height, width))
orangeErrors = np.reshape(orangeErrors, (height, width))
greenErrors = np.reshape(greenErrors, (height, width))
for i in range(height):
for j in range(width):
yellowError = yellowErrors[i,j]
orangeError = orangeErrors[i,j]
greenError = greenErrors[i,j]
if (yellowError > 11 and orangeError > 14 and greenError > 12.5):
continue
elif (yellowError == min(yellowError, orangeError, greenError)):
yellowMask[i,j] = 255
elif (orangeError == min(yellowError, orangeError, greenError)):
orangeMask[i,j] = 255
elif (greenError == min(yellowError, orangeError, greenError)):
greenMask[i,j] = 255
yellowMask = cv2.erode(yellowMask, None, iterations=1)
yellowMask = cv2.dilate(yellowMask, None, iterations=2)
orangeMask = cv2.erode(orangeMask, None, iterations=1)
orangeMask = cv2.dilate(orangeMask, None, iterations=2)
greenMask = cv2.erode(greenMask, None, iterations=1)
greenMask = cv2.dilate(greenMask, None, iterations=2)
yellowContours, hierarchy = cv2.findContours(yellowMask, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
orangeContours, hierarchy = cv2.findContours(orangeMask, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
greenContours, hierarchy = cv2.findContours(greenMask, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
if (len(yellowContours) != 0):
maxContour = max(yellowContours, key = cv2.contourArea)
center, radius = cv2.minEnclosingCircle(maxContour)
cv2.circle(image, (int(center[0]), int(center[1])), int(radius), \
color=(0, 255, 255), thickness=2)
cv2.circle(image, (int(center[0]), int(center[1])), 1, \
color=(0, 0, 255), thickness=1)
#cv2.drawContours(image, [maxContour], contourIdx=-1, color=(0, 255, 255), thickness=2)
if (len(orangeContours) != 0):
maxContour = max(orangeContours, key = cv2.contourArea)
center, radius = cv2.minEnclosingCircle(maxContour)
cv2.circle(image, (int(center[0]), int(center[1])), int(radius), \
color=(0, 125, 255), thickness=2)
cv2.circle(image, (int(center[0]), int(center[1])), 1, \
color=(0, 0, 255), thickness=1)
#cv2.drawContours(image, [maxContour], contourIdx=-1, color=(0, 125, 255), thickness=2)
if (len(greenContours) != 0):
maxContour = max(greenContours, key = cv2.contourArea)
center, radius = cv2.minEnclosingCircle(maxContour)
cv2.circle(image, (int(center[0]), int(center[1])), int(radius), \
color=(0, 255, 0), thickness=2)
cv2.circle(image, (int(center[0]), int(center[1])), 1, \
color=(0, 0, 255), thickness=1)
#cv2.drawContours(image, [maxContour], contourIdx=-1, color=(0, 255, 0), thickness=2)
return image
def runApplication(self, videoFile, saveVideo=False):
# Create video stream object
videoCapture = cv2.VideoCapture(videoFile)
# Define video codec and output file if video needs to be saved
if (saveVideo == True):
fourcc = cv2.VideoWriter_fourcc(*'mp4v')
# 720p 30fps video
out = cv2.VideoWriter('BuoyDetection.mp4', fourcc, 30, (1280, 720))
# Continue to process frames if the video stream object is open
while(videoCapture.isOpened()):
ret, frame = videoCapture.read()
# Continue processing if a valid frame is received
if ret == True:
newFrame = self.detectBuoys(frame)
# Save video if desired, resizing frame to 720p
if (saveVideo == True):
out.write(cv2.resize(newFrame, (1280, 720)))
# Display frame to the screen in a video preview
cv2.imshow("Frame", cv2.resize(newFrame, (1280, 720)))
# Exit if the user presses 'q'
if cv2.waitKey(1) & 0xFF == ord('q'):
break
# If the end of the video is reached, wait for final user keypress and exit
else:
cv2.waitKey(0)
break
# Release video and file object handles
videoCapture.release()
if (saveVideo == True):
out.release()
print('Video and file handles closed')
kmeans_obj = KMeans(3, x) kmeans_obj.fit(3, 0.002) means = kmeans_obj.mean_vec cov_mat_list = kmeans_obj.CovMatrix() mixture_coeff = kmeans_obj.MixtureCoeff() print(cov_mat_list) """from sklearn.cluster import KMeans obj = KMeans(n_clusters = 3, init = 'k-means++', max_iter = 100, n_init = 10, random_state = 0) y_Kmeans = obj.fit_predict(x) print(obj.cluster_centers_[:])""" GMM_obj = GMM(3, x, means, cov_mat_list, mixture_coeff) GMM_obj.fit(0.0002) print(GMM_obj.mean_vec) print(GMM_obj.cov_mat) print(GMM_obj.mixture_coeff) y_pred = GMM_obj.ClusterPredict(x) plt.scatter(GMM_obj.x_train[y_pred == 0, 0], GMM_obj.x_train[y_pred == 0, 1], s = 20, c = 'red', label = 'Cluster 1') plt.scatter(GMM_obj.x_train[y_pred == 1, 0], GMM_obj.x_train[y_pred == 1, 1], s = 20, c = 'green', label = 'Cluster 2') plt.scatter(GMM_obj.x_train[y_pred == 2, 0], GMM_obj.x_train[y_pred == 2, 1], s = 20, c = 'blue', label = 'Cluster 3') plt.scatter(GMM_obj.mean_vec[:, 0], GMM_obj.mean_vec[:, 1], s = 50, c = 'yellow', label = 'Centroids') plt.show() plt.scatter(GMM_obj.x_train[:, 0], GMM_obj.x_train[:, 1]) plt.show()
def xai_feature(self, samp_num, option= 'None'):
"""extract the important features from the input data
Arg:
fea_num: number of features that needed by the user
samp_num: number of data used for explanation
return:
fea: extracted features
"""
print '----------------------------------------------------'
print "parameters:"
print "data:",self.data
print "data shape:", self.data.shape
print "seq_len:", self.seq_len
print "start:", self.start
print "sp:", self.sp
print "real_sp:", self.real_sp
print "pred:", self.pred
print "trunc_len:", self.tl
print "trunc_data",self.trunc_data
print "trunc_data_test", self.trunc_data_test
print '----------------------------------------------------'
cen = self.seq_len/2
half_tl = self.tl/2
sample = np.random.randint(1, self.tl+1, samp_num)
print "sample:",sample
features_range = range(self.tl+1)
data_explain = np.copy(self.trunc_data).reshape(1, self.trunc_data.shape[0])
data_sampled = np.copy(self.trunc_data_test)
for i, size in enumerate(sample, start=1):
inactive = np.random.choice(features_range, size, replace=False)
#print '\ninactive --->',inactive
tmp_sampled = np.copy(self.trunc_data)
tmp_sampled[inactive] = 0
#tmp_sampled[inactive] = np.random.choice(range(257), size, replace = False)
#print "trunc_data.shape", self.trunc_data.shape
tmp_sampled = tmp_sampled.reshape(1, self.trunc_data.shape[0])
data_explain = np.concatenate((data_explain, tmp_sampled), axis=0)
#print "data_explain.shape", data_explain.shape
data_sampled_mutate = np.copy(self.data)
if self.real_sp < half_tl:
data_sampled_mutate[0, 0:tmp_sampled.shape[1]] = tmp_sampled
elif self.real_sp >= self.seq_len - half_tl:
data_sampled_mutate[0, (self.seq_len - tmp_sampled.shape[1]): self.seq_len] = tmp_sampled
else:
data_sampled_mutate[0, (self.real_sp - half_tl):(self.real_sp + half_tl + 1)] = tmp_sampled
data_sampled = np.concatenate((data_sampled, data_sampled_mutate),axis=0)
if option == "Fixed":
print "Fix start points"
data_sampled[:, self.real_sp] = self.start
label_sampled = self.model.predict(data_sampled, verbose = 0)[:, self.real_sp, 1]
label_sampled = label_sampled.reshape(label_sampled.shape[0], 1)
#X = r.matrix(data_explain, nrow = data_explain.shape[0], ncol = data_explain.shape[1])
#Y = r.matrix(label_sampled, nrow = label_sampled.shape[0], ncol = label_sampled.shape[1])
#n = r.nrow(X)
#print "n:", n
#p = r.ncol(X)
#print "p:", p
#print "np.sqrt(n*np.log(p)):", np.sqrt(n*np.log(p))
#print "X_shape", X.dim
#print "Y_shape", Y.dim
#Mixture model fitting
gmm = GMM(label_sampled,n_components=2).fit(data_explain)
print gmm.converged_
means = gmm.means_
covariances = gmm.covariances_
r_ik = np.zeros((samp_num+1,2))
k=-1
for m,c in zip(means,covariances):
k += 1
reg_cov = 5e-5*np.identity(self.tl+1)
c = c + reg_cov
#print "C:", c
multi_normal = multivariate_normal(mean=m,cov=c)
r_ik[:,k] = gmm.weights_[k] * multi_normal.pdf(data_explain)
#mat_norm = np.zeros((501,501))
#np.fill_diagonal(mat_norm, 1/np.sum(r_ik,axis=1))
#P = mat_norm.dot(r_ik)
res = np.argmax(r_ik, axis=1)
# find the index for the best component
best_component_idx = res[0]
# fitting beta according to best component of mixture regression model
# get the data for this component
idx=np.where(res==best_component_idx)[0]
X = r.matrix(data_explain[idx], nrow = len(idx), ncol = self.tl+1)
Y = r.matrix(label_sampled[idx], nrow = len(idx), ncol = 1)
n = r.nrow(X)
print "n:", n
p = r.ncol(X)
print "p:", p
print "np.sqrt(n*np.log(p)):", np.sqrt(n*np.log(p))
# solve fused lasso by r library and get the importance score from the results
print "X_shape", X.dim
print "Y_shape", Y.dim
results = r.fusedlasso1d(y=Y, X=X)
#print "result_i", result_i
result = np.array(r.coef(results, np.sqrt(n*np.log(p)))[0])[:,-1]
print "result:", result
#results = r.fusedlasso1d(y=Y,X=X)
#result = np.array(r.coef(results, np.sqrt(n*np.log(p)))[0])[:,-1]
# sorting the importance_score and return the important features
importance_score = np.argsort(result)[::-1]
print 'importance_score ...',importance_score
self.fea = (importance_score-self.tl/2)+self.real_sp
self.fea = self.fea[np.where(self.fea<200)]
self.fea = self.fea[np.where(self.fea>=0)]
print 'self.fea ...',self.fea
return self.fea
BETA = 1
# set random seed to reproduce
tf.random.set_seed(SEED)
np.random.seed(SEED)
#Get data
log = logger()
label_name = os.path.join("data", "labels")
data_name = os.path.join("data", "data_set")
labels = log.unpickle(label_name)
labels = np.expand_dims(labels, axis=-1)
data_set = log.unpickle(data_name)
#Initialize the models and optimizers
gmm_model = GMM(data_set.shape[1], K=K)
optimizer = tf.keras.optimizers.Adam(
learning_rate=0.0001) #Low learning rate needed!
#train function
@tf.function #--> optimizes the program by making a graph
def train_models(data, label):
with tf.GradientTape(persistent=True) as tape:
sample, prob, mean, logvar = gmm_model(label)
log_likelihood = gmm_model.log_likelihood(data, prob, mean, logvar)
grad = tape.gradient(log_likelihood, gmm_model.variables)
optimizer.apply_gradients(zip(grad, gmm_model.variables))
del tape
return log_likelihood
plt.plot(x2[:, 0], x2[:, 1], 'o')
plt.plot(np.mean(x2[:, 0]), np.mean(x2[:, 1]), 'x', color='black')
plt.text(np.mean(x2[:, 0]), np.mean(x2[:, 1]), '$\mu_{e}$')
plt.xlabel('x')
plt.ylabel('y')
plt.title('Labelled data and corresponding means')
plt.show()
M = 2
max_iter = 200
tol = 1e-3
diagonal = False
gmm = GMM(X, M)
# run the K-means algorithm firstly to initialize the means of the GMM algorithm
# mu_0 = random.sample(list(X), M)
# mu_0, D = k_means(X, M, mu_0=mu_0, max_iter=max_iter, tol=tol, interactive=False)
# 1.) EM algorithm for GMM:
# TODO
L = gmm.EM(max_iter=max_iter, tol=tol, interactive=False, diagonal=False)
plt.ioff()
plt.plot(L)
plt.xlabel('Iteration')
plt.ylabel('Value')
plt.title('EM log-likelihood function')
plt.show()
CURL_TRAIN_SIZE = 50 * N_CENTERS
CURL_TEST_SIZE = 1000
BATCH_SIZE = 100
# Model
INPUT_DIM = DIM
HIDDEN_DIM = 20
OUT_DIM = HIDDEN_DIM
# Training
N_EPOCH = 1000
LR = 1e-3
# Data generation
CENTERS = torch.randn(N_CENTERS * DIM).view(N_CENTERS, DIM)
gmm = GMM(DIM, CENTERS, VARIANCE)
X_train, y_train = gmm.sample(TRAIN_SAMPLES)
X_test, y_test = gmm.sample(TEST_SAMPLES)
train_CURL = ContrastiveDataset(
*build_CURL_dataset(X_train, y_train, CURL_TRAIN_SIZE))
assert len(train_CURL) == CURL_TRAIN_SIZE
test_CURL = ContrastiveDataset(
*build_CURL_dataset(X_test, y_test, CURL_TEST_SIZE))
train_data = GMMDataset(X_train, y_train)
test_data = GMMDataset(X_test, y_test)
train_loader = DataLoader(train_data, shuffle=True, batch_size=BATCH_SIZE)
test_loader = DataLoader(test_data, shuffle=False, batch_size=BATCH_SIZE)
def main():
Parser = argparse.ArgumentParser()
Parser.add_argument('datagenModule', type=str)
Parser.add_argument('algName', type=str)
#TODO:Parser.add_argument( 'modelName', type=str )
Parser.add_argument('-K', '--K', type=int, default=3)
Parser.add_argument('--alpha0', type=float, default=1.0)
Parser.add_argument('--covar_type', type=str, default='full')
Parser.add_argument('--min_covar', type=float, default=1e-9)
# Batch learning args
Parser.add_argument('--nIter', type=int, default=100)
# Online learning args
Parser.add_argument('--batch_size', type=int, default=100)
Parser.add_argument('--nBatch', type=int, default=50)
Parser.add_argument('--nRep', type=int, default=1)
Parser.add_argument('--rhoexp', type=float, default=0.5)
Parser.add_argument('--rhodelay', type=float, default=1)
# Generic args
Parser.add_argument('--jobname', type=str, default='defaultjob')
Parser.add_argument('--taskid', type=int, default=1)
Parser.add_argument('--nTask', type=int, default=1)
Parser.add_argument('--initname', type=str, default='random')
Parser.add_argument('--seed', type=int, default=8675309)
Parser.add_argument('-v',
'--doVerbose',
action='store_true',
default=False)
Parser.add_argument('--printEvery', type=int, default=5)
Parser.add_argument('--saveEvery', type=int, default=10)
Parser.add_argument('--doProfile', action='store_true', default=False)
args = Parser.parse_args()
modelParams = dict()
for argName in ['K', 'covar_type', 'min_covar', 'alpha0']:
modelParams[argName] = args.__getattribute__(argName)
dataParams = dict()
for argName in ['nBatch', 'nRep', 'batch_size', 'seed']:
dataParams[argName] = args.__getattribute__(argName)
algParams = dict()
for argName in ['initname', 'nIter', 'rhoexp', 'rhodelay', \
'nIter', 'printEvery', 'saveEvery']:
algParams[argName] = args.__getattribute__(argName)
# Dynamically load module provided by user as data-generator
# this must implement a generator function called "minibatch_generator" or "get_data"
datagenmod = __import__('GMM.data.' + args.datagenModule,
fromlist=['GMM', 'data'])
if 'print_data_info' in dir(datagenmod):
datagenmod.print_data_info()
gmm = GMM.GMM(**modelParams)
gmm.print_model_info()
for task in xrange(args.taskid, args.taskid + args.nTask):
basepath = os.path.join('results', args.algName, args.jobname,
str(task))
mkpath(basepath)
algParams['savefilename'] = os.path.join(basepath, 'trace')
seed = hash(args.jobname + str(task)) % np.iinfo(int).max
algParams['seed'] = seed
print 'Trial %2d/%d | savefile: %s | seed: %d' % (
task, args.nTask, algParams['savefilename'], algParams['seed'])
if args.algName.startswith('o') and args.algName.count('EM') > 0:
DataGen = datagenmod.minibatch_generator(**dataParams)
gmm = GMM.GMM(**modelParams)
em = OEM.OnlineEMLearnerGMM(gmm, **algParams)
em.fit(DataGen, seed)
elif args.algName.count('sklearnEM') > 0:
sklgmm = sklearn.mixture.GMM( n_components=args.K, random_state=seed, covariance_type=args.covar_type, \
min_covar=args.min_covar, n_init=1, n_iter=args.nIter, init_params='' )
X = datagenmod.get_data(**dataParams)
gmm = GMM.GMM(**modelParams)
em = EM.EMLearnerGMM(gmm, **algParams)
em.init_params(X, seed=seed)
sklgmm.weights_ = gmm.w
sklgmm.means_ = gmm.mu
sklgmm.covars_ = gmm.Sigma
sklgmm.fit(X)
elif args.algName.count('EM') > 0:
Data = datagenmod.get_data(**dataParams)
gmm = GMM.GMM(**modelParams)
em = EM.EMLearnerGMM(gmm, **algParams)
em.fit(Data, seed)
elif args.algName.count('DPVB') > 0:
Data = datagenmod.get_data(**dataParams)
D = Data.shape[1]
gw = GaussWishDistr.GaussWishDistr(D=D)
qdp = QDPGMM.QDPGMM(gw, **modelParams)
em = VB.VBLearnerGMM(qdp, **algParams)
em.fit(Data, seed)
elif args.algName.count('VB') > 0:
Data = datagenmod.get_data(**dataParams)
D = Data.shape[1]
dF = D + 1
invW = np.eye(D)
gw = GaussWishDistr.GaussWishDistr(D=D)
qgmm = QGMM.QGMM(gw, **modelParams)
em = VB.VBLearnerGMM(qgmm, **algParams)
em.fit(Data, seed)
from GMM import GMM
from sklearn import mixture
# generate the dataset
X, Y = make_classification(n_samples=1000,
n_features=2,
n_redundant=0,
n_informative=2,
n_clusters_per_class=2)
X = preprocessing.scale(X)
num_clusters = 3
num_epochs = 50
gmm_model = GMM()
phi, pi_dist, mean, covariance = gmm_model.fit(X,
num_clusters=num_clusters,
num_epochs=num_epochs)
gmm_sklearn = mixture.GaussianMixture(n_components=2)
gmm_sklearn.fit(X)
plt.figure(figsize=(8, 8))
plt.subplots_adjust(left=0.05, bottom=0.05, right=0.95, top=0.9)
plt.subplot(211)
plt.title('Plot for the unclustered data', fontsize='small')
plt.scatter(X[:, 0], X[:, 1], s=25, c=None)
plt.subplot(212)
plt.title('Plot for the clustered data', fontsize='small')
i = np.random.randint(0, nSamples-1, nComp)
means = samples[i]
covars = np.empty((nComp, 3), np.float32)
covars[:] = 10
gmm_cpu = mixture.GMM(nComp)
gmm_cpu.dtype = np.float32
gmm_cpu.init_params = ''
gmm_cpu.means_ = means
gmm_cpu.weights_ = weights
gmm_cpu.covars_ = covars
gmm_cpu.fit(samples)
gmm = GMM(context, nIter, nComp, nSamples)
a = calcA_cpu(weights, means, covars)
cl.enqueue_copy(queue, gmm.dA, a).wait()
gmm.has_preset_wmc = True
w,m,c = gmm.fit(dSamples, nSamples, retParams=True)
print 'converged: {0}'.format(gmm.has_converged)
print gmm_cpu.weights_
print w
print
print gmm_cpu.means_
print m
print
print gmm_cpu.covars_
# 2. 再分类
#=======================================================================
from GMM import GMM
import numpy as np
from sklearn.datasets import make_moons
from sklearn.model_selection import train_test_split
import matplotlib.pyplot as plt
from util import *
# 构造聚类数据,X是特征数据，Y是相应的label，此时生成的是半环形图
X, Y = make_moons(n_samples=1000, noise=0.04, random_state=0)
# 划分数据，一部分用于训练聚类，一部分用于分类
X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=0.2)
model = GMM(X_train, K=10)
# 获取各个类别的概率
result = model.fit()
print('每条数据属于各个类别的概率如下: ', result)
# 获取每条数据所在的类别
label_train = np.argmax(result, axis=1)
print(label_train)
# 获取测试数据所在的类别的概率
result_test = model.predict(X_test)
# 获取测试数据的类别
label_test = np.argmax(result_test, axis=1)
# 展示原始数据分布及其label
ax1 = plt.subplot(211)
import numpy as np
import matplotlib.pyplot as plt
from GMM import GMM
if __name__ == '__main__':
group_a = np.random.normal(loc=(20.00, 14.00), scale=(4.0, 4.0), size=(1000, 2))
group_b = np.random.normal(loc=(15.00, 8.00), scale=(2.0, 2.0), size=(1000, 2))
group_c = np.random.normal(loc=(30.00, 40.00), scale=(2.0, 2.0), size=(1000, 2))
group_d = np.random.normal(loc=(25.00, 32.00), scale=(7.0, 7.0), size=(1000, 2))
data = np.concatenate((group_a, group_b, group_c, group_d))
g = GMM(n_components=4)
eval_train = g.train(data)
for c in g.components:
print '*****'
print c.mean
print c.cov
plt.plot(eval_train)
plt.show()
# estimate bandwidth for mean shift
bandwidth = cluster.estimate_bandwidth(X, quantile=params['quantile'])
# connectivity matrix for structured Ward
connectivity = kneighbors_graph(X,
n_neighbors=params['n_neighbors'],
include_self=False)
# make connectivity symmetric
connectivity = 0.5 * (connectivity + connectivity.T)
# ============
# 初始化所有聚类算法
# ============
# 自编的K-Means、GMM算法
my_kmeans = K_Means(n_clusters=params['n_clusters'])
my_gmm = GMM(n_clusters=params['n_clusters'])
my_spec = Spectral()
# sklearn中自带的算法
ms = cluster.MeanShift(bandwidth=bandwidth, bin_seeding=True)
two_means = cluster.MiniBatchKMeans(n_clusters=params['n_clusters'])
ward = cluster.AgglomerativeClustering(n_clusters=params['n_clusters'],
linkage='ward',
connectivity=connectivity)
spectral = cluster.SpectralClustering(n_clusters=params['n_clusters'],
eigen_solver='arpack',
affinity="nearest_neighbors")
dbscan = cluster.DBSCAN(eps=params['eps'])
optics = cluster.OPTICS(min_samples=params['min_samples'],
xi=params['xi'],
min_cluster_size=params['min_cluster_size'])
affinity_propagation = cluster.AffinityPropagation(
class QuickBrush(Brush):
lWorksize = (16, 16)
def __init__(self, context, devices, d_img, d_labels):
Brush.__init__(self, context, devices, d_labels)
self.context = context
self.queue = cl.CommandQueue(context,
properties=cl.command_queue_properties.PROFILING_ENABLE)
nComponentsFg = 4
nComponentsBg = 4
self.nDim = 3
self.dim = d_img.dim
filename = os.path.join(os.path.dirname(__file__), 'quick.cl')
program = createProgram(context, context.devices, [], filename)
# self.kernSampleBg = cl.Kernel(program, 'sampleBg')
self.kern_get_samples = cl.Kernel(program, 'get_samples')
self.lWorksize = (16, 16)
self.gWorksize = roundUp(self.dim, self.lWorksize)
nSamples = 4 * (self.gWorksize[0] / self.lWorksize[0]) * (
self.gWorksize[1] / self.lWorksize[1])
# self.gmmFg_cpu = mixture.GMM(4)
self.gmmFg = GMM(context, 65, nComponentsFg, 10240)
self.gmmBg = GMM(context, 65, nComponentsBg, nSamples)
self.hScore = np.empty(self.dim, np.float32)
self.hSampleFg = np.empty((10240, ), np.uint32)
self.hSampleBg = np.empty((12000, ), np.uint32)
self.hA = np.empty((max(nComponentsFg, nComponentsBg), 8), np.float32)
self.d_img = d_img
cm = cl.mem_flags
self.dSampleFg = cl.Buffer(context, cm.READ_WRITE, size=4 * 10240)
self.dSampleBg = cl.Buffer(context, cm.READ_WRITE, size=4 * 12000)
self.dA = cl.Buffer(context, cm.READ_ONLY | cm.COPY_HOST_PTR, hostbuf=self.hA)
self.dScoreFg = Buffer2D(context, cm.READ_WRITE, self.dim, np.float32)
self.dScoreBg = Buffer2D(context, cm.READ_WRITE, self.dim, np.float32)
#self.points = Set()
self.capPoints = 200 * 200 * 300 #brush radius 200, stroke length 300
self.points = np.empty((self.capPoints), np.uint32)
# self.colorize = Colorize.Colorize(clContext, clContext.devices)
# self.hTriFlat = self.hTri.reshape(-1)
# self.probBg(1200)
self.h_img = np.empty(self.dim, np.uint32)
self.h_img = self.h_img.ravel()
cl.enqueue_copy(self.queue, self.h_img, self.d_img, origin=(0, 0), region=self.dim).wait()
self.samples_bg_idx = np.random.randint(0, self.dim[0] * self.dim[1], 12000)
self.hSampleBg = self.h_img[self.samples_bg_idx]
cl.enqueue_copy(self.queue, self.dSampleBg, self.hSampleBg).wait()
w,m,c = self.gmmBg.fit(self.dSampleBg, 300, retParams=True)
print w
print m
print c
self.gmmBg.score(self.d_img, self.dScoreBg)
pass
def draw(self, p0, p1):
Brush.draw(self, p0, p1)
#self.probFg(x1-20, x1+20, y1-20, y1+20)
#return
"""color = self.colorTri[self.type]
#self.argsScore[5] = np.int32(self.nComponentsFg)
#seed = []
hasSeeds = False
redoBg = False
minX = sys.maxint
maxX = -sys.maxint
minY = sys.maxint
maxY = -sys.maxint
for point in self.points[0:nPoints]:
#if self.hTriFlat[point] != color:
self.hTriFlat[point] = color
#seed += point
hasSeeds = True
minX = min(minX, point%self.width)
maxX = max(maxX, point%self.width)
minY = min(minY, point/self.width)
maxY = max(maxY, point/self.width)
#if (point[1]*self.width + point[0]) in self.randIdx:
# redoBg = True
#if redoBg:
# self.probBg(0)
#if len(seed) == 0:
if not hasSeeds:
return
minX = max(0, minX-DILATE)
maxX = min(self.width-1, maxX + DILATE)
minY = max(0, minY-DILATE)
maxY = min(self.height-1, maxY + DILATE)
"""
args = [
np.int32(self.n_points),
self.d_points,
cl.Sampler(self.context, False, cl.addressing_mode.NONE,
cl.filter_mode.NEAREST),
self.d_img,
self.dSampleFg
]
gWorksize = roundUp((self.n_points, ), (256, ))
self.kern_get_samples(self.queue, gWorksize, (256,), *args).wait()
cl.enqueue_copy(self.queue, self.hSampleFg, self.dSampleFg)
# print self.hSampleFg.view(np.uint8).reshape(10240, 4)[0:self.n_points, :]
# print self.n_points
self.gmmFg.fit(self.dSampleFg, self.n_points)
# print w
# print m
# print c
self.gmmFg.score(self.d_img, self.dScoreFg)
# self.argsSampleBg = [
# self.d_labels,
# np.int32(self.label),
# cl.Sampler(self.context, False, cl.addressing_mode.NONE,
# cl.filter_mode.NEAREST),
# self.d_img,
# self.dSampleFg
# ]
#
# gWorksize = roundUp(self.dim, (16, 16))
#
# self.kernSampleBg(self.queue, gWorksize, (16, 16),
# *(self.argsSampleBg)).wait()
# cl.enqueue_copy(self.queue, self.hSampleBg, self.dSampleBg).wait()
pass
def probFg(self, d_samples, n_points):
# if True:
# tri = self.hTri[minY:maxY, minX:maxX]
# b = (tri == self.colorTri[self.type])
#
# samplesFg = self.hSrc[minY:maxY, minX:maxX]
# samplesFg = samplesFg[b]
# else:
# DILATE = 5
# samplesFg = self.hSrc[minY:maxY, minX:maxX].ravel()
#gpu = False
#self.prob(self.gmmFG, samplesFg, self.dScoreFg, gpu)
#self.gmmFg_cpu.fit(samplesFg)
#print 'cpu', self.gmmFg_cpu.weights_
#a = calcA_cpu(self.gmmFg_cpu.weights_.astype(np.float32), self.gmmFg_cpu.means_.astype(np.float32), self.gmmFg_cpu.covars_.astype(np.float32))
#cl.enqueue_copy(self.queue, self.gmmFg.dA, a).wait()
#weights, means, covars = self.gmmFg.fit(samplesFg, retParams=True)
#a = calcA_cpu(weights, means[:, 0:3], covars[:, 0:3])
#cl.enqueue_copy(self.queue, self.gmmFg.dA, a).wait()
w,m,c = self.gmmFg.fit(d_samples, n_points, retParams=True)
print w
print m
print c
#print 'gpu', weights
self.gmmFg.score(self.d_img, self.dScoreFg)
#score returns float64, not float32 -> convert with astype
#self.hScore = -self.gmmFG.score(self.rgb.reshape(-1, 3)).astype(np.float32)
"""
def drawCircle(self, xc, yc, points=None):
r = self.radius
for y in xrange(-r, r):
for x in xrange(-r, r):
if points != None:
points.add((xc+x, yc+y))
"""
def probBg(self, nSamples):
#self.kernSampleBg(self.queue, self.gWorksize, self.lWorksize, *(self.argsSampleBg)).wait()
#cl.enqueue_copy(self.queue, self.hSampleBg, self.dSampleBg).wait()
self.bgIdx = np.where(self.hTri.ravel() != self.colorTri[self.type])[0]
self.randIdx = self.bgIdx[np.random.randint(0, len(self.bgIdx), 2000)]
self.bgIdx = np.setdiff1d(self.bgIdx, self.randIdx)
self.hSampleBg[0:len(self.randIdx)] = self.hSrc.view(np.uint32).ravel()[
self.randIdx]
cl.enqueue_copy(self.queue, self.dSampleBg, self.hSampleBg).wait()
#print self.gmmBg.fit(self.hSrc.view(np.uint32).ravel()[self.randIdx], retParams=True)
self.gmmBg.fit(self.hSrc.view(np.uint32).ravel()[self.randIdx])
#self.gmmBg.fit(self.dSampleBg, nSamples=len(self.randIdx))
self.gmmBg.score(self.dSrc, self.dScoreBg)
#!/usr/bin/env python3 # -*- coding: utf-8 -*- """ Created on Sun Apr 21 02:43:24 2019 @author: maachou """ from sklearn.datasets.samples_generator import make_blobs import matplotlib.pyplot as plt from GMM import GMM mix=GMM(K=6) X,Y = make_blobs(cluster_std=0.5,random_state=20,n_samples=100,centers=6) plt.scatter(X[:,0],X[:,1]) print(X.shape) mix.fit(X) mix.Means() Y=mix.predict(X) plt.scatter(X[:,0],X[:,1],c=Y)
def __init__(self, context, devices, d_img, d_labels):
Brush.__init__(self, context, devices, d_labels)
self.context = context
self.queue = cl.CommandQueue(context,
properties=cl.command_queue_properties.PROFILING_ENABLE)
nComponentsFg = 4
nComponentsBg = 4
self.nDim = 3
self.dim = d_img.dim
filename = os.path.join(os.path.dirname(__file__), 'quick.cl')
program = createProgram(context, context.devices, [], filename)
# self.kernSampleBg = cl.Kernel(program, 'sampleBg')
self.kern_get_samples = cl.Kernel(program, 'get_samples')
self.lWorksize = (16, 16)
self.gWorksize = roundUp(self.dim, self.lWorksize)
nSamples = 4 * (self.gWorksize[0] / self.lWorksize[0]) * (
self.gWorksize[1] / self.lWorksize[1])
# self.gmmFg_cpu = mixture.GMM(4)
self.gmmFg = GMM(context, 65, nComponentsFg, 10240)
self.gmmBg = GMM(context, 65, nComponentsBg, nSamples)
self.hScore = np.empty(self.dim, np.float32)
self.hSampleFg = np.empty((10240, ), np.uint32)
self.hSampleBg = np.empty((12000, ), np.uint32)
self.hA = np.empty((max(nComponentsFg, nComponentsBg), 8), np.float32)
self.d_img = d_img
cm = cl.mem_flags
self.dSampleFg = cl.Buffer(context, cm.READ_WRITE, size=4 * 10240)
self.dSampleBg = cl.Buffer(context, cm.READ_WRITE, size=4 * 12000)
self.dA = cl.Buffer(context, cm.READ_ONLY | cm.COPY_HOST_PTR, hostbuf=self.hA)
self.dScoreFg = Buffer2D(context, cm.READ_WRITE, self.dim, np.float32)
self.dScoreBg = Buffer2D(context, cm.READ_WRITE, self.dim, np.float32)
#self.points = Set()
self.capPoints = 200 * 200 * 300 #brush radius 200, stroke length 300
self.points = np.empty((self.capPoints), np.uint32)
# self.colorize = Colorize.Colorize(clContext, clContext.devices)
# self.hTriFlat = self.hTri.reshape(-1)
# self.probBg(1200)
self.h_img = np.empty(self.dim, np.uint32)
self.h_img = self.h_img.ravel()
cl.enqueue_copy(self.queue, self.h_img, self.d_img, origin=(0, 0), region=self.dim).wait()
self.samples_bg_idx = np.random.randint(0, self.dim[0] * self.dim[1], 12000)
self.hSampleBg = self.h_img[self.samples_bg_idx]
cl.enqueue_copy(self.queue, self.dSampleBg, self.hSampleBg).wait()
w,m,c = self.gmmBg.fit(self.dSampleBg, 300, retParams=True)
print w
print m
print c
self.gmmBg.score(self.d_img, self.dScoreBg)
pass
def fit(self, X, K, eps=pow(10, -2)):
# fits the parameters of the HMM using EM algorithm
# X is the sequence of observations (array of size (T,D)),
# K is the number of hidden states
# eps : tolerance on log likelihood difference between two iterations for convergence of EM algorithm
self.K = K
T, D = X.shape
# initialization of means and covariances with GMM
print(
"Initialization of Gaussians parameters (means and covariances) with GMM : "
)
gmm_model = GMM(isotropic=False)
gmm_model.fit(X, K, eps=eps)
self.mus = gmm_model.mus
self.Sigmas2 = gmm_model.Sigmas2
print("\nFit of HMM : ")
# initialization of pis and A at random
self.pis = np.random.rand(self.K)
self.pis /= np.sum(self.pis)
self.A = np.random.rand(self.K, self.K)
self.A /= np.sum(self.A, axis=1)[:, None]
lik = self.compute_log_likelihood(X)
print("Initial log-likelihood : ", lik)
delta_lik = 1
cpt_iter = 1
while (delta_lik > eps):
# Expectation step
pi = self.compute_proba_Zt_cond_X(
X) # array (T,K) (t,i) -> p(z_t = i|X; θ)
pij = self.compute_proba_Zt_and_Znext_cond_X(
X) # tensor (T-1,K,K) (t,i,j) -> p(z_(t+1) = j, z t = i|X; θ)
# Maximization step
self.pis = pi[0, :]
pi_repeated = pi[:, :, np.newaxis] # (T,K,D)
self.mus = np.sum(pi_repeated * X[:, np.newaxis, :],
axis=0) / np.sum(pi_repeated, axis=0)
self.Sigmas2 = []
for k in range(self.K):
Xc = X - self.mus[k]
Sigmas2k = 0
for t in range(T):
xt = Xc[t, :][:, None] # size (d,1)
Sigmas2k += np.dot(xt, xt.T) * pi[t, k]
Sigmas2k /= np.sum(pi[:, k])
self.Sigmas2.append(Sigmas2k)
self.Sigmas2 = np.array(self.Sigmas2)
self.A = np.sum(pij, axis=0) / np.sum(pi[:-1], axis=0)[:, None]
# Computing new likelihood, and deciding if we should stop
old_lik = lik # storing old_likelihood to compute delta_lik
lik = self.compute_log_likelihood(X) # storing new likelihood
delta_lik = lik - old_lik # measure to decide if we should stop or iterate again
print("Iter " + str(cpt_iter) + " ; log_likelihood : " + str(lik))
cpt_iter += 1
print("EM algorithm converged.")
print("initial distribution found (rounded, 2 decimals) : ",
np.round(self.pis, 2))
print("transition matrix found (rounded, 2 decimals) : ",
np.round(self.A, 2))
def test_data(length, array_length):
#txtName = "causal_continue_noise_0.4_normal_sample_1000_length_200.txt"
#f = file(txtName, "a+")
counter11 = 0
counter10 = 0
counter01 = 0
counter00 = 0
counter11_01 = 0
counter10_01 = 0
counter01_01 = 0
counter00_01 = 0
counter_undecided = 0
counter_true = 0
counter_false = 0
counter_undecided2 = 0
counter_true2 = 0
counter_false2 = 0
counter_error_1 = 0
counter_error_2 = 0
p_array_granger1 = []
p_array_granger2 = []
p_array_CUTE1 = []
p_array_CUTE2 = []
p_array_improve_CUTE1 = []
p_array_improve_CUTE2 = []
p_array1 = []
p_array2 = []
p_array_granger = []
for i in range(0, 1000):
write_str = ""
p = random.randint(1, 3)
#effect, test1 = generate_continue_data(200, p)
#cause, effect = generate_continue_data(150, p)
#cause_tmp = list(cause)
#effect_tmp = list(effect)
#cause = zero_change(cause)
#effect = zero_change(effect)
#cause,effect = generate_continue_data_with_change_lag(350,10)
cause = GMM(3, array_length)
effect = GMM(5, array_length)
cause_tmp = list(cause)
effect_tmp = list(effect)
#effect = forward_shift_continue_data(cause,p)
#noise = np.random.normal(0, 0.1, 200)
#for j in range(0, 200):
# effect[j] = effect[j] + noise[j]
#for i in range(0,len(cause)):
#cause[i]=math.tanh(cause[i])
#cause[i] = math.pow(math.e,cause[i])
#effect[i] = math.pow(math.e,effect[i])
#cause[i] = math.pow(cause[i],3)/10
#effect[i] = math.pow(effect[i],3)/10
#effect[i]=math.tanh(effect[i])
#effect[i] = math.pow(effect[i],3)
#effect = GMM(8,200)
#plt.plot(cause)
#plt.plot(effect)
#plt.show()
#cause = normalize(cause)
#effect = normalize(effect)
#cause = normalize_data(cause)
#effect = normalize_data(effect)
#cause = zero_change(cause)
#effect = zero_change(effect)
from scipy.special import expit
#for i in range(0,len(effect)):
#effect[i]=expit(effect[i])
#effect[i] = 1.0/effect[i]
for ii in range(0, len(cause)):
write_str = write_str + " " + str(cause[ii])
for jj in range(0, len(effect)):
write_str = write_str + " " + str(effect[jj])
#print "cause:" + str(cause)
#print "effect:" + str(effect)
# effect, test2 = ge_normal_data(p,200)
print "Continuous data, Granger causality test"
print "cause->effect"
p_value_cause_to_effect1 = []
flag1 = False
#ce1 = grangercausalitytests([[effect[i], cause[i]] for i in range(0, len(cause))], p)
ce_p = granger(cause, effect, -1)
#for key in ce1:
# p_value_cause_to_effect1.append(ce1[key][0]["params_ftest"][1])
# if ce1[key][0]["params_ftest"][1] < 0.05:
# flag1 = True
if ce_p < 0.05:
flag1 = True
print "effect->cause"
p_value_effect_to_cause2 = []
flag2 = False
#ce2 = grangercausalitytests([[cause[i], effect[i]] for i in range(0, len(cause))], p)
ce2_p = granger(effect, cause, -1)
#for key in ce2:
# p_value_effect_to_cause2.append(ce2[key][0]["params_ftest"][1])
# if ce2[key][0]["params_ftest"][1] < 0.05:
# flag2 = True
if ce2_p < 0.05:
flag2 = True
if ce_p < 0.05:
p_array_granger1.append(ce_p)
elif ce2_p < 0.05:
p_array_granger2.append(ce2_p)
if flag1 and flag2:
print "Continuous data，Granger two-way cause and effect"
write_str = write_str + " " + "连续数据，格兰杰双向因果"
counter11 += 1
elif flag1 and not flag2:
print "Continuous data，Granger correct cause and effect"
write_str = write_str + " " + "连续数据，格兰杰正确因果"
counter10 += 1
p_array_granger.append(ce_p)
elif not flag1 and flag2:
print "Continuous data，Granger wrong cause and effect"
write_str = write_str + " " + "连续数据，格兰杰错误因果"
counter01 += 1
elif not flag1 and not flag2:
print "Continuous data，Granger no cause and effect"
write_str = write_str + " " + "连续数据，格兰杰没有因果"
#break
counter00 += 1
#write_str = write_str + " " + str(min(p_value_cause_to_effect1)) + " " + str(min(p_value_effect_to_cause2))
cause2 = get_type_array(cause, length)
effect2 = get_type_array(effect, length)
print "01 data, Granger causality test"
print "cause->effect"
p_value_cause_to_effect3 = []
flag3 = False
#ce3 = grangercausalitytests([[effect2[i], cause2[i]] for i in range(0, len(cause2))], p)
ce3_p = granger(cause2, effect2, -1)
#for key in ce3:
# p_value_cause_to_effect3.append(ce3[key][0]["params_ftest"][1])
# if ce3[key][0]["params_ftest"][1] < 0.05:
# flag3 = True
if ce3_p < 0.05:
flag3 = True
print "effect->cause"
p_value_effect_to_cause4 = []
flag4 = False
#ce4 = grangercausalitytests([[cause2[i], effect2[i]] for i in range(0, len(cause2))], p)
ce4_p = granger(effect2, cause2, -1)
#for key in ce4:
# p_value_effect_to_cause4.append(ce4[key][0]["params_ftest"][1])
# if ce4[key][0]["params_ftest"][1] < 0.05:
# flag4 = True
if ce4_p < 0.05:
flag4 = True
if flag3 and flag4:
print "01 data，Granger two-way cause and effect"
write_str = write_str + " " + "离散数据，格兰杰双向因果"
counter11_01 += 1
elif flag3 and not flag4:
print "01 data，Granger correct cause and effect"
write_str = write_str + " " + "离散数据，格兰杰正确因果"
counter10_01 += 1
elif not flag3 and flag4:
print "01 data，Granger wrong cause and effect"
write_str = write_str + " " + "离散数据，格兰杰错误因果"
counter01_01 += 1
elif not flag3 and not flag4:
print "01 data，Granger no cause and effect"
write_str = write_str + " " + "离散数据，格兰杰没有因果"
counter00_01 += 1
#write_str = write_str + " " + str(min(p_value_cause_to_effect3)) + " " + str(min(p_value_effect_to_cause4))
print
delta_ce = calculate_difference3(cause, effect, length)
delta_ec = calculate_difference3(effect, cause, length)
print 'cause' + ' -> ' + 'effect' + ':' + str(delta_ce)
print 'effect' + ' -> ' + 'cause' + ':' + str(delta_ec)
if delta_ce > delta_ec and delta_ce - delta_ec >= -math.log(0.05, 2):
print "CUTE，correct cause and effect"
write_str = write_str + " " + "CUTE，正确因果"
counter_true += 1
elif delta_ec > delta_ce and delta_ec - delta_ce >= -math.log(0.05, 2):
print "CUTE，wrong cause and effect"
write_str = write_str + " " + "CUTE，错误因果"
counter_false += 1
else:
print "CUTE，undecided"
write_str = write_str + " " + "CUTE，未决定"
counter_undecided += 1
write_str = write_str + " " + str(pow(2, -abs(delta_ce - delta_ec)))
p = math.pow(2, -(delta_ce - delta_ec))
if p < 1:
p_array1.append(p)
else:
p_array2.append(math.pow(2, -(delta_ec - delta_ce)))
#f.write(write_str)
#f.write("\n")
cause = change_to_zero_one(cause_tmp)
effect = change_to_zero_one(effect_tmp)
cause2effect = bernoulli2(effect, length) - cbernoulli2(
effect, cause, length)
effect2cause = bernoulli2(cause, length) - cbernoulli2(
cause, effect, length)
# print 'cause' + ' -> ' + 'effect' + ':' + str(cause2effect)
# print 'effect' + ' -> ' + 'cause' + ':' + str(effect2cause)
p = math.pow(2, -(cause2effect - effect2cause))
if p < 1:
p_array_improve_CUTE1.append(p)
else:
p_array_improve_CUTE2.append(
math.pow(2, -(effect2cause - cause2effect)))
cause2effect = bernoulli(effect) - cbernoulli(effect, cause)
effect2cause = bernoulli(cause) - cbernoulli(cause, effect)
if p < 1:
p_array_CUTE1.append(p)
else:
p_array_CUTE2.append(math.pow(2, -(effect2cause - cause2effect)))
print
print "*****************************cut line*****************************"
print
#f.close()
print "连续数据，格兰杰因果关系检验："
print "双向因果:" + str(counter11)
print "正确因果:" + str(counter10)
print "错误因果:" + str(counter01)
print "没有因果" + str(counter00)
print "-----------------"
print "离散数据，格兰杰因果关系检验："
print "双向因果:" + str(counter11_01)
print "正确因果:" + str(counter10_01)
print "错误因果:" + str(counter01_01)
print "没有因果" + str(counter00_01)
print "-----------------"
print "discret data，snml causality test："
print "correct cause and effect:" + str(counter_true)
print "wrong cause and effect:" + str(counter_false)
print "no cause and effect:" + str(counter_undecided)
print "-----------------"
print "01 data，CUTE causality test："
granger_test = (bh_procedure(p_array_granger1, 0.05) +
bh_procedure(p_array_granger2, 0.05)) / 1000.0
ourmodel = (bh_procedure(p_array1, 0.05) +
bh_procedure(p_array2, 0.05)) / 1000.0
cute = (bh_procedure(p_array_CUTE1, 0.05) +
bh_procedure(p_array_CUTE2, 0.05)) / 1000.0
improve_cute = (bh_procedure(p_array_improve_CUTE1, 0.05) +
bh_procedure(p_array_improve_CUTE2, 0.05)) / 1000.0
print granger_test
print improve_cute
print ourmodel
return granger_test, ourmodel, cute, improve_cute
