def generate_data(choice='linear'):
global X, Y, N
if choice == 'linear':
N = np.random.randint(50, 150)
std = (np.random.randint(6, 10)) * 0.1
X, Y = make_blobs(n_samples=N,
centers=2,
random_state=0,
cluster_std=0.60)
elif choice == 'circle':
N = np.random.randint(100, 200)
factor = (np.random.randint(1, 6)) * 0.1
X, Y = make_circles(n_samples=N, factor=factor, noise=0.1)
plt.delaxes()
plt.scatter(X[:, 0],
X[:, 1],
c=Y,
cmap='winter',
s=100,
edgecolors='black')
plt.xlabel("Feature1")
plt.ylabel("Feature2")
plt.title("DATA POINTS")
gendat = BytesIO()
plt.savefig(gendat, format='png')
# figfile.seek(0) # Rewind to the beginning of the file
gendat_img = base64.b64encode(gendat.getbuffer()).decode('ascii')
return gendat_img
def circlesexample(self):
X, y = make_circles(90, factor=0.2, noise=0.1)
r = np.exp(-(X**2).sum(1))
zaxis = [0.2, 0.4, 0.6, 0.8, 1.0]
zaxislabel = [r'0.2', r'0.4', r'0.6', r'0.8', r'1.0']
self.ax.scatter(X[:, 0], X[:, 1], r, c=y, s=70, cmap='seismic')
self.ax.view_init(elev=90, azim=90)
self.ax.set_xlabel('X',
color='w',
fontproperties=self.prop,
fontsize=60)
self.ax.set_ylabel('Y',
color='w',
fontproperties=self.prop,
fontsize=60)
self.ax.set_zlabel('Z',
labelpad=-1,
color='red',
fontproperties=self.prop,
fontsize=60)
self.ax.set_zticklabels(zaxislabel, fontsize=7, color='none')
# self.ax.set_zticks([], False)
self.ax.set_zticks(zaxis)
plt.xticks(ticks=np.arange(-1.2, 1.4, .2), labels='')
plt.yticks(ticks=np.arange(-1.2, 1.4, .2), labels='')
self.ax.grid(linewidth=20)
return self.fig,
def __init__(self):
self.start_time = time.time()
if (Configuration.data == "Social_Network_Ads.csv"):
self.dataset = pd.read_csv(str(Configuration.data))
if (Configuration.algorithm == "linear_regression"):
self.X = self.dataset.iloc[:, :-1].values
self.y = self.dataset.iloc[:, 1].values
elif (Configuration.algorithm == "logistic_regression" or Configuration.algorithm == "svc"
or Configuration.algorithm == "decision_tree_classification" or Configuration.algorithm == "random_forest_classification" or
Configuration.algorithm == "knn"):
if (Configuration.data=="Social_Network_Ads.csv"):
self.X = self.dataset.iloc[:, [2,3]].values
self.y = self.dataset.iloc[:, 4].values
else:
if (Configuration.data == "moons"):
from sklearn.datasets.samples_generator import make_moons
self.X, self.y = make_moons(100, noise=.2, random_state = 0)
elif (Configuration.data == "circles"):
from sklearn.datasets.samples_generator import make_circles
self.X, self.y = make_circles(100, factor=.5, noise=.1, random_state = 0)
elif (Configuration.algorithm == "polynomial_regression"):
self.X = self.dataset.iloc[:, 1:2].values
self.y = self.dataset.iloc[:, 2].values
elif (Configuration.algorithm == "kmeans"):
self.X = self.dataset.iloc[:, [3, 4]].values
self.y = None
if (Configuration.data == "Social_Network_Ads.csv"):
self.directory = "SocialNetworkAds"
elif (Configuration.data == "moons"):
self.directory = "Moons"
elif (Configuration.data == "circles"):
self.directory = "Circles"
def ex2():
X, y = make_circles(100, factor=.1, noise=.1)
clf = SVC(kernel='linear').fit(X, y)
plt.scatter(X[:, 0], X[:, 1], c=y, s=50, cmap='summer')
plt.show()
plt.scatter(X[:, 0], X[:, 1], c=y, s=50, cmap='summer')
plot_svc_decision_function(clf);
plt.show()
r = np.exp(-(X[:, 0] ** 2 + X[:, 1] ** 2))
ax = plt.subplot(projection='3d')
ax.scatter3D(X[:, 0], X[:, 1], r, c=y, s=50, cmap='summer')
ax.view_init(elev=30, azim=30)
ax.set_xlabel('x')
ax.set_ylabel('y')
ax.set_zlabel('r')
plt.show()
clf = SVC(kernel='rbf')
clf.fit(X, y)
plt.scatter(X[:, 0], X[:, 1], c=y, s=50, cmap='summer')
plot_svc_decision_function(clf)
plt.scatter(clf.support_vectors_[:, 0], clf.support_vectors_[:, 1],
s=200, facecolors='none');
plt.show()
pass
def make_circles(self, ax=None, plot=False):
if ax is None:
fig, ax = plt.subplots(1, 1, figsize=(12, 6))
X, y = make_circles(100, factor=.1, noise=.1)
if plot:
mask = y > 0
ax.scatter(X[mask, 0],
X[mask, 1],
c=self.val_to_color(y[mask]),
s=50,
label="Positive")
ax.scatter(X[~mask, 0],
X[~mask, 1],
c=self.val_to_color(y[~mask]),
s=50,
label="Negative")
# plt.scatter(X[:, 0], X[:, 1], c=self.val_to_color(y), s=50, cmap='autumn')
ax.set_xlabel("$x_1$")
ax.set_ylabel("$x_2$")
ax.legend()
return X, y
def main(_):
cut = int(FLAGS.n_samples * 0.7)
start = time.time()
data, features = make_circles(n_samples=FLAGS.n_samples, shuffle=True, noise=0.12, factor=0.4)
tr_data, tr_features = data[:cut], features[:cut]
te_data, te_features = data[cut:], features[cut:]
test = []
fig, ax = plt.subplots()
ax.scatter(tr_data[:, 0], tr_data[:, 1],
marker='o', s=100, c=tr_features, cmap=plt.cm.coolwarm)
plt.plot()
plt.show()
with tf.Session() as sess:
for i, j in zip(te_data, te_features):
distances = tf.reduce_sum(tf.square(tf.subtract(i, tr_data)), axis=1)
neighbor = tf.arg_min(distances, 0)
test.append(tr_features[sess.run(neighbor)])
fig, ax = plt.subplots()
ax.scatter(te_data[:, 0], te_data[:, 1],
marker='o', s=100, c=test, cmap=plt.cm.coolwarm)
plt.plot()
plt.show()
end = time.time()
print("Found in %.2f seconds" % (end-start))
print("Cluster assignments:", test)
def kernel_model_rbf():
x, y = make_circles(100, factor=.1, noise=.1)
clf = SVC(kernel='rbf', C=1E6).fit(x, y)
plt.scatter(x[:, 0], x[:, 1], c=y, s=50, cmap='autumn')
plot_svc_decision_function(clf, plot_support=False)
plot_3d(x, y)
plt.show()
def get_toy_classification_data(n_samples=100, centers=3, n_features=2, type_data = "blobs"):
# generate 2d classification dataset
if (type_data == "blobs"):
X, y = make_blobs(n_samples=n_samples, centers=centers, n_features=n_features)
elif(type_data == "moons"):
X, y = make_moons(n_samples=n_samples, noise=0.1)
elif(type_data == "circles"):
X, y = make_circles(n_samples=n_samples, noise=0.05)
# scatter plot, dots colored by class value
# df = DataFrame(dict(x=X[:,0], y=X[:,1], label=y))
# colors = {0:'red', 1:'blue', 2:'green'}
# fig, ax = pyplot.subplots()
# grouped = df.groupby('label')
# for key, group in grouped:
# group.plot(ax=ax, kind='scatter', x='x', y='y', label=key, color=colors[key])
# pyplot.show()
X_train, X_test, y_train, y_test = train_test_split( X, y, test_size=0.25, stratify = None)
classes = np.unique(y_train)
if(0):
enc = OneHotEncoder().fit(classes.reshape(-1,1))
y_train = enc.transform(y_train.reshape(-1, 1))
print (y_test)
y_test = enc.transform(y_test.reshape(-1, 1))
print (y_test)
y_train = one_hot_encode(y_train, classes)
y_test = one_hot_encode(y_test, classes)
return X_train, y_train, X_test, y_test, classes
def _get_circles(*args, **kwargs):
X, y = make_circles(n_samples=100, noise=0.1)
metadata = {
'regression': False,
'scoring': classifier_scoring,
'primary_metric': 'accuracy',
}
return X, y, metadata
def make_circles(self, plot=False):
X, y = make_circles(100, factor=.1, noise=.1)
if plot:
plt.scatter(X[:, 0], X[:, 1], c=y, s=50, cmap='autumn')
plt.xlabel("$x_1$")
plt.ylabel("$x_2$")
return X, y
def generate_circle_data(N=100, seed=1):
np.random.seed(seed)
X, y = make_circles(N, factor=.1, noise=.1)
y[y == 0] = -1
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.33,
random_state=42)
return X_train.tolist(), X_test.tolist(), y_train.tolist(), y_test.tolist()
def kernel_model_linear():
"""
线性核函数
:return:
"""
x, y = make_circles(100, factor=.1, noise=.1)
clf = SVC(kernel='linear').fit(x, y)
plt.scatter(x[:, 0], x[:, 1], c=y, s=50, cmap='autumn')
plot_svc_decision_function(clf, plot_support=False)
plot_3d(x, y)
plt.show()
def spectral_clustering(g_directed):
# X=np.array.g_directed.edges()
# W = pairwise_distances(X, metric="euclidean")
# vectorizer = np.vectorize(lambda x: 1 if x < 5 else 0)
# W = np.vectorize(vectorizer)(W)
# print(W)
W = nx.adjacency_matrix(g_directed)
print(W.todense())
D = np.diag(np.sum(np.array(W.todense()), axis=1))
print('degree matrix:')
print(D)
L = D - W
print('laplacian matrix:')
print(L)
e, v = np.linalg.eig(L)
# eigenvalues
print('eigenvalues:')
print(e)
# eigenvectors
print('eigenvectors:')
print(v)
i = np.where(e < 0.5)[0]
x = 1
U = np.array(v[:, i[1]])
km = KMeans(init='k-means++', n_clusters=3)
km.fit(U)
km.labels_
X, clusters = make_circles(n_samples=1000,
noise=.05,
factor=.5,
random_state=0)
plt.scatter(X[:, 0], X[:, 1])
# Using K-means
km = KMeans(init='k-means++', n_clusters=2)
km_clustering = km.fit(X)
plt.scatter(X[:, 0],
X[:, 1],
c=km_clustering.labels_,
cmap='rainbow',
alpha=0.7,
edgecolors='b')
# Using Spectral Clustering scitkit-learn’s implementation
sc = SpectralClustering(n_clusters=2,
affinity='nearest_neighbors',
random_state=0)
sc_clustering = sc.fit(X)
plt.scatter(X[:, 0],
X[:, 1],
c=sc_clustering.labels_,
cmap='rainbow',
alpha=0.7,
edgecolors='b')
def nonLinearSeparable():
from sklearn.datasets.samples_generator import make_circles
X, y = make_circles(100, factor=0.1, noise=.1)
plt.scatter(X[:, 0], X[:, 1], c=y, s=50, cmap='autumn')
plt.show()
from sklearn import svm
clf = svm.SVC(kernel='rbf', C=1E6)
clf.fit(X, y)
plt.scatter(X[:, 0], X[:, 1], c=y, s=50, cmap='autumn')
plot_decision_function(clf)
def get_dataset(self):
if self.type == 'moon':
datas, labels = make_moons(n_samples=self.n_samples,
noise=self.noise)
elif self.type == 'circle':
datas, labels = make_circles(n_samples=self.n_samples,
noise=self.noise)
else:
print('wrong dataset type input.')
dataset = {}
dataset['data'] = datas
dataset['target'] = labels
return dataset
def train_svm_plus():
# 二维圆形数据 factor 内外圆比例（0， 1）
X, y = make_circles(100, factor=0.1, noise=0.1)
# 加入径向基函数
clf = SVC(kernel='rbf')
clf.fit(X, y)
plt.scatter(X[:, 0], X[:, 1], c=y, s=50, cmap='autumn')
plot_SVC_decision_function(clf, plot_support=False)
plt.scatter(clf.support_vectors_[:, 0],
clf.support_vectors_[:, 1],
s=300,
lw=1,
facecolors='none')
return X, y
def test_mapper_filterer():
data, labels = make_circles(n_samples=2000, noise=0.03, factor=0.3)
params = {
'coverer__intervals': 10,
'coverer__overlap': 0.1,
'clusterer__min_samples': 3,
'clusterer__eps': 0.5
}
m_filter = Mapper(filterer=MinMaxScaler(), params=params)
m_nofilter = Mapper(filterer=MinMaxScaler(), params=params)
scaled_data = MinMaxScaler().fit_transform(data)
m_filter.fit(data)
m_nofilter.fit(data, scaled_data)
assert_true(m_filter.links_ == m_nofilter.links_)
assert_true(len(m_filter.nodes_) == len(m_nofilter.nodes_))
def single_cluster(dim, samples, std, clusters, cluster_func, x=None, y=None):
# x, y = make_blobs(n_samples=samples, centers=clusters, n_features=dim, random_state=0, cluster_std=std)
if x is None or y is None:
x, y = samples_generator.make_circles(n_samples=samples,
random_state=True,
factor=0.3,
noise=0.05)
_y = cluster_func(x, clusters)
acc = sklearn.metrics.homogeneity_score(y, _y)
hyp.plot(x, '.', group=y, save_path='or.png')
hyp.plot(x, '.', group=_y, save_path='grp.png')
print('Accuracy {0:0.2f}'.format(acc))
def test_graph_simple():
data, labels = make_circles(n_samples=2000, noise=0.03, factor=0.3)
params = {
'coverer__intervals': 10,
'coverer__overlap': 0.1,
'clusterer__min_samples': 3,
'clusterer__eps': 0.5
}
m = Mapper(params=params)
scaled_data = MinMaxScaler().fit_transform(data)
m.fit(data, scaled_data)
categories = {"labels": labels}
scales = {"y[0]": scaled_data[:, 0], "y[1]": scaled_data[:, 1]}
json_graph_str = json_graph(m, categories, scales)
# check if it can be loaded to validate html
json_graph_dict = json.loads(json_graph_str)
html_graph_str = html_graph(m, categories, scales) # validate HTML?
def choose_dataset(chosen_dataset, n_points):
X = None
if chosen_dataset == "blobs":
X = make_blobs(n_samples=n_points,
centers=4,
n_features=2,
cluster_std=1.5,
random_state=42)[0]
elif chosen_dataset == "moons":
X = make_moons(n_samples=n_points, noise=0.05, random_state=42)[0]
elif chosen_dataset == "scatter":
X = make_blobs(n_samples=n_points,
cluster_std=[2.5, 2.5, 2.5],
random_state=42)[0]
elif chosen_dataset == "circle":
X = make_circles(n_samples=n_points, noise=0, random_state=42)[0]
return X
def main():
# 过滤警告
warnings.filterwarnings("ignore")
# 创建“点滴”数据
# x, y = samples_generator.make_blobs(n_samples=200, centers=2, cluster_std=1, random_state=0)
# 创建“月牙”数据
# x, y = samples_generator.make_moons(n_samples=200, noise=0.05, random_state=0)
# 创建“环形”数据
x, y = samples_generator.make_circles(n_samples=200, noise=0.05, random_state=0, factor=0.4)
"""
创建七种聚类算法
"""
# clusters = cluster.KMeans(2) # K-means++
# clusters = cluster.MeanShift() # 均值迁移
# clusters = cluster. AgglomerativeClustering(2) # 层聚类
# clusters = cluster.AffinityPropagation() # AP聚类
# clusters = cluster.SpectralClustering(n_clusters=2, affinity="nearest_neighbors") # 谱聚类
# clusters = cluster.DBSCAN(eps=0.55, min_samples=5) # 密度聚类
clusters = GaussianMixture(n_components=2) # 高斯分布
# 拟合
_x = clusters.fit_predict(x)
"""
三种评价方法
"""
# 1.轮廓系数
print(metrics.silhouette_score(x, _x))
# 2.CH指数
print(metrics.calinski_harabasz_score(x, _x))
# 3.戴维森堡丁指数
print(metrics.davies_bouldin_score(x, _x))
# 绘图
plt.scatter(x[:, 0], x[:, 1], c=_x, cmap='viridis')
plt.show()
import numpy as np
import matplotlib.pyplot as plt
from sklearn.datasets.samples_generator import make_circles, make_blobs
from sklearn.cluster import DBSCAN
from sklearn.metrics import calinski_harabaz_score
# Create cluster dataset
X1, y1 = make_circles(n_samples=5000, factor=0.6, noise=0.05, random_state=666)
X2, y2 = make_blobs(n_samples=1000, n_features=2, centers=[[1.2, 1.2]], \
cluster_std=[[0.1]], random_state=666)
X = np.concatenate((X1, X2))
y = np.concatenate((y1, y2))
# Create cluster model
dbscan = DBSCAN(eps=0.1, min_samples=5)
y_predict = dbscan.fit_predict(X)
print('Calinski-Harabasz Index Score: ', calinski_harabaz_score(X, y_predict))
# Visualization
plt.scatter(X[:, 0], X[:, 1], c=y_predict, marker='o', edgecolors='black')
plt.xlabel('X')
plt.ylabel('y')
plt.title('DBSCAN Cluster Algorithm')
plt.show()
import time
import matplotlib
import matplotlib.pyplot as plt
from sklearn.datasets.samples_generator import make_circles
N = 210
K = 2
MAX_ITERS = 1000
cut = int(N * 0.7)
start = time.time()
data, features = make_circles(n_samples=N,
shuffle=True,
noise=0.12,
factor=0.4)
# print(data.shape)
# print(data)
# print(features.shape)
tr_data, tr_features = data[:cut], features[:cut]
te_data, te_features = data[cut:], features[cut:]
test = []
fig, ax = plt.subplots()
ax.scatter(tr_data.transpose()[0],
tr_data.transpose()[1],
marker='o',
s=100,
c=tr_features,
import matplotlib
import matplotlib.pyplot as plt
from sklearn.datasets.samples_generator import make_blobs
from sklearn.datasets.samples_generator import make_circles
DATA_TYPE = 'blobs'
# Number of clusters, if we choose circles, only 2 will be enough
if (DATA_TYPE == 'circle'):
K = 2
else:
K = 4
# Maximum number of iterations, if the conditions are not met
MAX_ITERS = 1000
N = 200
start = time.time()
centers = [(-2, -2), (-2, 1.5), (1.5, -2), (2, 1.5)]
if (DATA_TYPE == 'circle'):
data, features = make_circles(n_samples=200, shuffle=True, noise=0.01, factor=0.4)
else:
data, features = make_blobs(n_samples=200, centers=centers, n_features=2, cluster_std=0.8, shuffle=False, random_state=42)
fig, ax = plt.subplots()
ax.scatter(np.asarray(centers).transpose()[0], np.asarray(centers).transpose()[1], marker = 'o', s = 250)
plt.show()
print("TEST")
def main(selectedDataset="digits", pop_size=100, max_generations=100):
# a few hard-coded values
figsize = [5, 4]
seed = 42
# pop_size = 300
offspring_size = 2 * pop_size
# max_generations = 300
maximize = False
# selectedDataset = "circles"
selectedClassifiers = ["SVC"]
# a list of classifiers
allClassifiers = [
[RandomForestClassifier, "RandomForestClassifier", 1],
[BaggingClassifier, "BaggingClassifier", 1],
[SVC, "SVC", 1],
[RidgeClassifier, "RidgeClassifier", 1],
# [AdaBoostClassifier, "AdaBoostClassifier", 1],
# [ExtraTreesClassifier, "ExtraTreesClassifier", 1],
# [GradientBoostingClassifier, "GradientBoostingClassifier", 1],
# [SGDClassifier, "SGDClassifier", 1],
# [PassiveAggressiveClassifier, "PassiveAggressiveClassifier", 1],
# [LogisticRegression, "LogisticRegression", 1],
]
selectedClassifiers = [classifier[1] for classifier in allClassifiers]
folder_name = datetime.datetime.now().strftime(
"%Y-%m-%d-%H-%M") + "-archetypes-" + selectedDataset + "-" + str(
pop_size)
if not os.path.exists(folder_name):
os.makedirs(folder_name)
else:
sys.stderr.write("Error: folder \"" + folder_name +
"\" already exists. Aborting...\n")
sys.exit(0)
# open the logging file
logfilename = os.path.join(folder_name, 'logfile.log')
logger = setup_logger('logfile_' + folder_name, logfilename)
logger.info("All results will be saved in folder \"%s\"" % folder_name)
# load different datasets, prepare them for use
logger.info("Preparing data...")
# synthetic databases
centers = [[1, 1], [-1, -1], [1, -1]]
blobs_X, blobs_y = make_blobs(n_samples=400,
centers=centers,
n_features=2,
cluster_std=0.6,
random_state=seed)
circles_X, circles_y = make_circles(n_samples=400,
noise=0.15,
factor=0.4,
random_state=seed)
moons_X, moons_y = make_moons(n_samples=400, noise=0.2, random_state=seed)
iris = datasets.load_iris()
digits = datasets.load_digits()
# forest_X, forest_y = loadForestCoverageType() # local function
mnist_X, mnist_y = loadMNIST() # local function
dataList = [
[blobs_X, blobs_y, 0, "blobs"],
[circles_X, circles_y, 0, "circles"],
[moons_X, moons_y, 0, "moons"],
[iris.data, iris.target, 0, "iris4"],
[iris.data[:, 2:4], iris.target, 0, "iris2"],
[digits.data, digits.target, 0, "digits"],
# [forest_X, forest_y, 0, "covtype"],
[mnist_X, mnist_y, 0, "mnist"]
]
# argparse; all arguments are optional
parser = argparse.ArgumentParser()
parser.add_argument(
"--classifiers",
"-c",
nargs='+',
help="Classifier(s) to be tested. Default: %s. Accepted values: %s" %
(selectedClassifiers[0], [x[1] for x in allClassifiers]))
parser.add_argument(
"--dataset",
"-d",
help="Dataset to be tested. Default: %s. Accepted values: %s" %
(selectedDataset, [x[3] for x in dataList]))
parser.add_argument("--pop_size",
"-p",
type=int,
help="EA population size. Default: %d" % pop_size)
parser.add_argument("--offspring_size",
"-o",
type=int,
help="Ea offspring size. Default: %d" % offspring_size)
parser.add_argument("--max_generations",
"-mg",
type=int,
help="Maximum number of generations. Default: %d" %
max_generations)
# finally, parse the arguments
args = parser.parse_args()
# a few checks on the (optional) inputs
if args.dataset:
selectedDataset = args.dataset
if selectedDataset not in [x[3] for x in dataList]:
logger.info(
"Error: dataset \"%s\" is not an accepted value. Accepted values: %s"
% (selectedDataset, [x[3] for x in dataList]))
sys.exit(0)
if args.classifiers != None and len(args.classifiers) > 0:
selectedClassifiers = args.classifiers
for c in selectedClassifiers:
if c not in [x[1] for x in allClassifiers]:
logger.info(
"Error: classifier \"%s\" is not an accepted value. Accepted values: %s"
% (c, [x[1] for x in allClassifiers]))
sys.exit(0)
if args.max_generations: max_generations = args.max_generations
if args.pop_size: pop_size = args.pop_size
if args.offspring_size: offspring_size = args.offspring_size
# TODO: check that min_points < max_points and max_generations > 0
# print out the current settings
logger.info("Settings of the experiment...")
logger.info("Fixed random seed: %d" % (seed))
logger.info("Selected dataset: %s; Selected classifier(s): %s" %
(selectedDataset, selectedClassifiers))
logger.info(
"Population size in EA: %d; Offspring size: %d; Max generations: %d" %
(pop_size, offspring_size, max_generations))
# create the list of classifiers
classifierList = [x for x in allClassifiers if x[1] in selectedClassifiers]
# pick the dataset
db_index = -1
for i in range(0, len(dataList)):
if dataList[i][3] == selectedDataset:
db_index = i
dbname = dataList[db_index][3]
X, y = dataList[db_index][0], dataList[db_index][1]
number_classes = np.unique(y).shape[0]
logger.info("Creating train/test split...")
from sklearn.model_selection import StratifiedKFold
skf = StratifiedKFold(n_splits=3, shuffle=True, random_state=seed)
listOfSplits = [split for split in skf.split(X, y)]
trainval_index, test_index = listOfSplits[0]
X_trainval, y_trainval = X[trainval_index], y[trainval_index]
X_test, y_test = X[test_index], y[test_index]
skf = StratifiedKFold(n_splits=3, shuffle=False, random_state=seed)
listOfSplits = [split for split in skf.split(X_trainval, y_trainval)]
train_index, val_index = listOfSplits[0]
X_train, y_train = X_trainval[train_index], y_trainval[train_index]
X_val, y_val = X_trainval[val_index], y_trainval[val_index]
logger.info(
"Training set: %d lines (%.2f%%); test set: %d lines (%.2f%%)" %
(X_train.shape[0],
(100.0 * float(X_train.shape[0] / X.shape[0])), X_test.shape[0],
(100.0 * float(X_test.shape[0] / X.shape[0]))))
# rescale data
scaler = StandardScaler()
sc = scaler.fit(X_train)
X = sc.transform(X)
X_trainval = sc.transform(X_trainval)
X_train = sc.transform(X_train)
X_val = sc.transform(X_val)
X_test = sc.transform(X_test)
for classifier in classifierList:
classifier_name = classifier[1]
# start creating folder name
experiment_name = os.path.join(
folder_name,
datetime.datetime.now().strftime("%Y-%m-%d-%H-%M") +
"-archetypes-evolution-" + dbname + "-" + classifier_name)
if not os.path.exists(experiment_name): os.makedirs(experiment_name)
logger.info("Classifier used: " + classifier_name)
start = time.time()
solutions, trainAccuracy, testAccuracy = evolveArchetypes(
X,
y,
X_train,
y_train,
X_test,
y_test,
classifier,
pop_size,
offspring_size,
max_generations,
number_classes=number_classes,
maximize=maximize,
seed=seed,
experiment_name=experiment_name)
end = time.time()
exec_time = end - start
# only candidates with all classes are considered
final_archive = []
for sol in solutions:
c = sol.candidate
c = np.array(c)
y_core = c[:, -1]
if len(set(y_core)) == number_classes:
final_archive.append(sol)
logger.info("Now saving final Pareto front in a figure...")
pareto_front_x = [f.fitness[0] for f in final_archive]
pareto_front_y = [f.fitness[1] for f in final_archive]
figure = plt.figure(figsize=figsize)
ax = figure.add_subplot(111)
ax.plot(pareto_front_x,
pareto_front_y,
"bo-",
label="Solutions in final archive")
ax.set_title("Optimal solutions")
ax.set_xlabel("Archetype set size")
ax.set_ylabel("Error")
ax.set_xlim([1, X_train.shape[0]])
ax.set_ylim([0, 0.4])
plt.tight_layout()
plt.savefig(
os.path.join(
experiment_name,
"%s_EvoArch_%s_pareto.png" % (dbname, classifier_name)))
plt.savefig(
os.path.join(
experiment_name,
"%s_EvoArch_%s_pareto.pdf" % (dbname, classifier_name)))
plt.close(figure)
figure = plt.figure(figsize=figsize)
ax = figure.add_subplot(111)
ax.plot(pareto_front_x,
pareto_front_y,
"bo-",
label="Solutions in final archive")
ax.set_title("Optimal solutions")
ax.set_xlabel("Archetype set size")
ax.set_ylabel("Error")
plt.tight_layout()
plt.savefig(
os.path.join(
experiment_name,
"%s_EvoArch_%s_pareto_zoom.png" % (dbname, classifier_name)))
plt.savefig(
os.path.join(
experiment_name,
"%s_EvoArch_%s_pareto_zoom.pdf" % (dbname, classifier_name)))
plt.close(figure)
# initial performance
X_err, testAccuracy, model, fail_points, y_pred = evaluate_core(
X_trainval,
y_trainval,
X_test,
y_test,
classifier[0],
cname=classifier_name,
SEED=seed)
X_err, trainAccuracy, model, fail_points, y_pred = evaluate_core(
X_trainval,
y_trainval,
X_trainval,
y_trainval,
classifier[0],
cname=classifier_name,
SEED=seed)
logger.info("Compute performances!")
logger.info("Elapsed time (seconds): %.4f" % (exec_time))
logger.info("Initial performance: train=%.4f, test=%.4f, size: %d" %
(trainAccuracy, testAccuracy, X_train.shape[0]))
# best solution
accuracy = []
for sol in final_archive:
c = sol.candidate
c = np.array(c)
X_core = c[:, :-1]
y_core = c[:, -1]
X_err, accuracy_val, model, fail_points, y_pred = evaluate_core(
X_core,
y_core,
X_val,
y_val,
classifier[0],
cname=classifier_name,
SEED=seed)
X_err, accuracy_train, model, fail_points, y_pred = evaluate_core(
X_core,
y_core,
X_train,
y_train,
classifier[0],
cname=classifier_name,
SEED=seed)
accuracy.append(np.mean([accuracy_val, accuracy_train]))
best_ids = np.array(np.argsort(accuracy)).astype('int')[::-1]
count = 0
for i in best_ids:
if count > 2:
break
c = final_archive[i].candidate
c = np.array(c)
X_core = c[:, :-1]
y_core = c[:, -1]
X_err, accuracy_train, model, fail_points, y_pred = evaluate_core(
X_core,
y_core,
X_train,
y_train,
classifier[0],
cname=classifier_name,
SEED=seed)
X_err, accuracy_val, model, fail_points, y_pred = evaluate_core(
X_core,
y_core,
X_val,
y_val,
classifier[0],
cname=classifier_name,
SEED=seed)
X_err, accuracy, model, fail_points, y_pred = evaluate_core(
X_core,
y_core,
X_test,
y_test,
classifier[0],
cname=classifier_name,
SEED=seed)
logger.info(
"Minimal train/val error: train: %.4f, val: %.4f; test: %.4f, size: %d"
% (accuracy_train, accuracy_val, accuracy, X_core.shape[0]))
if False: #(dbname == "mnist" or dbname == "digits") and count == 0:
if dbname == "mnist":
H, W = 28, 28
if dbname == "digits":
H, W = 8, 8
logger.info("Now saving figures...")
# save archetypes
for index in range(0, len(y_core)):
image = np.reshape(X_core[index, :], (H, W))
plt.figure()
plt.axis('off')
plt.imshow(image, cmap=plt.cm.gray_r)
plt.title('Label: %d' % (y_core[index]))
plt.tight_layout()
plt.savefig(
os.path.join(
experiment_name,
"digit_%d_idx_%d.pdf" % (y_core[index], index)))
plt.savefig(
os.path.join(
experiment_name,
"digit_%d_idx_%d.png" % (y_core[index], index)))
plt.close()
# save test errors
e = 1
for index in range(0, len(y_test)):
if fail_points[index] == True:
image = np.reshape(X_test[index, :], (H, W))
plt.figure()
plt.axis('off')
plt.imshow(image, cmap=plt.cm.gray_r)
plt.title('Label: %d - Prediction: %d' %
(y_test[index], y_pred[index]))
plt.savefig(
os.path.join(
experiment_name,
"err_lab_%d_pred_%d_idx_%d.pdf" %
(y_test[index], y_pred[index], e)))
plt.savefig(
os.path.join(
experiment_name,
"err_lab_%d_pred_%d_idx_%d.png" %
(y_test[index], y_pred[index], e)))
plt.close()
e = e + 1
# plot decision boundaries if we have only 2 dimensions!
if X.shape[1] == 2:
cmap = ListedColormap(sns.color_palette("bright", 3).as_hex())
xx, yy = make_meshgrid(X[:, 0], X[:, 1])
figure = plt.figure(figsize=figsize)
_, Z_0 = plot_contours(model, xx, yy, colors='k', alpha=0.2)
# plt.scatter(X_train[:, 0], X_train[:, 1], c=y_train, cmap=cmap, marker='s', alpha=0.4, label="train")
plt.scatter(X_test[:, 0],
X_test[:, 1],
c=y_test,
cmap=cmap,
marker='+',
alpha=0.3,
label="test")
plt.scatter(X_core[:, 0],
X_core[:, 1],
c=y_core,
cmap=cmap,
marker='D',
facecolors='none',
edgecolors='none',
alpha=1,
label="archetypes")
plt.scatter(X_err[:, 0],
X_err[:, 1],
marker='x',
facecolors='k',
edgecolors='k',
alpha=1,
label="errors")
plt.legend()
plt.title("%s - acc. %.4f" % (classifier_name, accuracy))
plt.tight_layout()
plt.savefig(
os.path.join(
experiment_name, "%s_EvoArch_%s_%d.png" %
(dbname, classifier_name, count)))
plt.savefig(
os.path.join(
experiment_name, "%s_EvoArch_%s_%d.pdf" %
(dbname, classifier_name, count)))
plt.close(figure)
if count == 0:
# using all samples in the training set
X_err, accuracy, model, fail_points, y_pred = evaluate_core(
X_trainval,
y_trainval,
X_test,
y_test,
classifier[0],
cname=classifier_name,
SEED=seed)
X_err_train, trainAccuracy, model_train, fail_points_train, y_pred_train = evaluate_core(
X_trainval,
y_trainval,
X_trainval,
y_trainval,
classifier[0],
cname=classifier_name,
SEED=seed)
figure = plt.figure(figsize=figsize)
_, Z_0 = plot_contours(model,
xx,
yy,
colors='k',
alpha=0.2)
plt.scatter(X_trainval[:, 0],
X_trainval[:, 1],
c=y_trainval,
cmap=cmap,
marker='s',
alpha=0.4,
label="train")
plt.scatter(X_test[:, 0],
X_test[:, 1],
c=y_test,
cmap=cmap,
marker='+',
alpha=0.4,
label="test")
plt.scatter(X_err[:, 0],
X_err[:, 1],
marker='x',
facecolors='k',
edgecolors='k',
alpha=1,
label="errors")
plt.legend()
plt.title("%s - acc. %.4f" % (classifier_name, accuracy))
plt.tight_layout()
plt.savefig(
os.path.join(
experiment_name, "%s_EvoArch_%s_alltrain.png" %
(dbname, classifier_name)))
plt.savefig(
os.path.join(
experiment_name, "%s_EvoArch_%s_alltrain.pdf" %
(dbname, classifier_name)))
plt.close(figure)
count = count + 1
logger.handlers.pop()
return
if q in oldCoreObjs.keys():
delte = [val for val in oldCoreObjs[q]
if val in notAccess] #Δ = N(q)∩Γ
queue.extend(delte) #将Δ中的样本加入队列Q
notAccess = [val for val in notAccess
if val not in delte] #Γ = Γ\Δ
k += 1
C[k] = [val for val in OldNotAccess if val not in notAccess]
for x in C[k]:
if x in coreObjs.keys():
del coreObjs[x]
return C
if __name__ == '__main__':
X, y_true = make_circles(n_samples=1000, noise=0.15) # 随机生成1000个圆环形状数据
print(X)
print(y_true)
plt.scatter(X[:, 0], X[:, 1], c=y_true)
plt.show()
# DBSCAN 算法
t0 = time.time()
y_pred = DBSCAN(eps=.1, min_samples=6).fit_predict(X) # 该算法对应的两个参数
t = time.time() - t0
plt.scatter(X[:, 0], X[:, 1], c=y_pred)
plt.title('time : %f' % t)
plt.show()
# eps为距离阈值ϵ，min_samples 为邻域样本数阈值MinPts, X为数据
ax = ax or plt.gca()
ax.scatter(X[:, 0], X[:, 1], c=y, s=50, cmap='autumn')
ax.set_xlim(-1, 4)
ax.set_ylim(-1, 6)
plot_svc_decision_function(model, ax)
fig, ax = plt.subplots(1, 2, figsize=(16, 6))
fig.subplots_adjust(left=0.0625, right=0.95, wspace=0.1)
for axi, N in zip(ax, [60, 120]):
plot_svm(N, axi)
axi.set_title('N = {0}'.format(N))
# 引入核函数的SVM
from sklearn.datasets.samples_generator import make_circles
X, y = make_circles(100, factor=.1, noise=.1)
clf = SVC(kernel='linear').fit(X, y)
plt.scatter(X[:, 0], X[:, 1], c=y, s=50, cmap='autumn')
plot_svc_decision_function(clf, plot_support=False)
plt.show()
#加入了新的维度r
from mpl_toolkits import mplot3d
r = np.exp(-(X**2).sum(1))
def plot_3D(elev=30, azim=30, X=X, y=y):
ax = plt.subplot(projection='3d')
ax.scatter3D(X[:, 0], X[:, 1], r, c=y, s=50, cmap='autumn')
ylim = ax.get_ylim()
xx = np.linspace(xlim[0], xlim[1], 200)
yy = np.linspace(ylim[0], ylim[1], 200)
YY, XX = np.meshgrid(yy, xx)
xy = np.vstack([XX.ravel(), YY.ravel()]).T
Z = model.decision_function(xy).reshape(XX.shape)
# plot decision boundary and margins
ax.contour(XX, YY, Z, colors='k', levels=[-1, 0, 1], alpha=0.5,
linestyles=['--', '-', '--'])
# plot support vectors
ax.scatter(model.support_vectors_[:, 0], model.support_vectors_[:, 1], s=300,
linewidth=1, facecolors='none', edgecolors='k')
plt.show()
y1_model = model.predict(X_train)
y2_model = model.predict(X_test)
print('Accuracy on train data:',accuracy_score(Y_train, y1_model))
print('Accuracy on test data:',accuracy_score(Y_test, y2_model))
X,Y=make_circles(n_samples=200,noise=0.05,factor=0.5)
X_train,X_test,Y_train,Y_test=train_test_split(X,Y,test_size=0.3)
svclassifier = svm.SVC(kernel='linear',C=1E3)
svclassifier.fit(X_train, Y_train)
plot_model(svclassifier)
print('accuracy test = ' ,accuracy_score(Y_test , svclassifier.predict(X_test)))
print('accuracy train = ' ,accuracy_score(Y_train , svclassifier.predict(X_train)))
P[i, j] = clf.decision_function([xi, yj])
# plot the margins
ax.contour(X, Y, P, colors='k',
levels=[-1, 0, 1], alpha=0.5,
linestyles=['--', '-', '--'])
plt.subplot(412)
plt.scatter(X[:, 0], X[:, 1], c=y, s=50, cmap='spring')
plot_svc_decision_function(clf)
plt.scatter(clf.support_vectors_[:, 0], clf.support_vectors_[:, 1],
s=200, facecolors='none')
"""
circles
"""
from sklearn.datasets.samples_generator import make_circles
X, y = make_circles(100, factor=.1, noise=.1)
clf = SVC(kernel='linear').fit(X, y)
plt.subplot(413)
plt.scatter(X[:, 0], X[:, 1], c=y, s=50, cmap='spring')
plot_svc_decision_function(clf)
#
clf = SVC(kernel='rbf')
clf.fit(X, y)
plt.subplot(414)
plt.scatter(X[:, 0], X[:, 1], c=y, s=50, cmap='spring')
plot_svc_decision_function(clf)
plt.scatter(clf.support_vectors_[:, 0], clf.support_vectors_[:, 1],
s=200, facecolors='none')
plt.show()
# -*- coding: utf-8 -*-
"""Demo108_PCA_Circles.ipynb
# **Tame Your Python**
"""
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import tensorflow as tf
print(tf.__version__)
from sklearn.datasets.samples_generator import make_circles
n = 100
# generate 2d classification dataset
X, y = make_circles(n_samples=n)
# scatter plot, dots colored by class value
df = pd.DataFrame(dict(x=X[:,0], y=X[:,1], label=y))
colors = {0:'red', 1:'blue'}
fig, ax = plt.subplots()
grouped = df.groupby('label')
for key, group in grouped:
group.plot(ax=ax, kind='scatter', x='x', y='y', label=key, color=colors[key])
plt.show()
datadict = {'X1': X[:,0],'X2' : X[:,1], 'target': y}
data = pd.DataFrame(data=datadict)
X = data.iloc[:,[0, 1]].values
type(X)
import numpy as np import matplotlib.pyplot as plt from sklearn.datasets.samples_generator import make_circles from sklearn.svm import SVC from DecisionBoundary import plot_svm_margin # Creating a toy data with circles X, y = make_circles(100, factor=.1, noise=.1, random_state=88) # plotting the data plt.figure(figsize=[6, 6]) plt.scatter(X[:, 0], X[:, 1], c=y, s=50, cmap=plt.cm.coolwarm) plt.show() # calculating the radius r = np.sum(X**2, axis=1)**0.5 # plotting the data, y-axis replaced with the radius plt.figure(figsize=[6, 6]) plt.scatter(X[:, 0], r, c=y, s=50, cmap=plt.cm.coolwarm) plt.show() # SVM R = np.vstack([X[:, 0], r]).T sv = SVC(kernel='linear', C=10000) sv.fit(R, y) # plotting the margin on the SVM of the transformed data plt.figure(figsize=[6, 6]) plot_svm_margin(R, y, sv) plt.show()
import time import matplotlib import matplotlib.pyplot as plt from sklearn.datasets.samples_generator import make_circles N=210 K=2 # Maximum number of iterations, if the conditions are not met MAX_ITERS = 1000 cut=int(N*0.7) start = time.time() data, features = make_circles(n_samples=N, shuffle=True, noise= 0.12, factor=0.4) tr_data, tr_features= data[:cut], features[:cut] te_data,te_features=data[cut:], features[cut:] fig, ax = plt.subplots() ax.scatter(tr_data.transpose()[0], tr_data.transpose()[1], marker = 'o', s = 100, c = tr_features, cmap=plt.cm.coolwarm ) plt.plot() points=tf.Variable(data) cluster_assignments = tf.Variable(tf.zeros([N], dtype=tf.int64)) sess = tf.Session() sess.run(tf.initialize_all_variables()) test=[]
import numpy as np
import matplotlib.pyplot as plt
from sklearn.datasets.samples_generator import make_circles
x, y = make_circles(
n_samples=1000,
noise=0.1,
factor=0.2,
random_state=0
)
#x.shape
# plt.figure(figsize=(5,5))
# #o = formato b = cor
# plt.plot(x[y==0,0],x[y==0,1],'ob',alpha=0.5)
# plt.plot(x[y==1,0],x[y==1,1],'xr',alpha=0.5)
# plt.xlim(-1.5,1.5)
# plt.ylim(-1.5,1.5)
# plt.legend(['0','1'])
# plt.title("Criar um grafico")
# #plt.show()
from keras.models import Sequential
from keras.layers import Dense
from keras.optimizers import SGD
model = Sequential()
model.add(Dense(4,input_shape=(2,),activation='tanh'))
model.add(Dense(1,activation='sigmoid'))
model.compile(SGD(lr=0.5),'binary_crossentropy',metrics=['accuracy'])
plot_svc_decision_function(clf)
plt.show()
# Note that a couple of the points touch the lines, these are known as our
# support vectors.
print clf.support_vectors_
# Visually check the concordance between the coordinates printed above
# and the highlighted points in the figure.
plt.scatter(x[:, 0], x[:, 1], c=y, s=50, cmap="spring")
plot_svc_decision_function(clf)
plt.scatter(clf.support_vectors_[:, 0], clf.support_vectors_[:, 1], s=200, facecolors="none")
plt.show()
# Only the support vectors matter with an SVM. Moving points wihtout letting
# them cross the decision boundaries, it would have no effect.
# The SVM becomes more powerful in conjunction with kernels. Let's look
# at some data which is not linearly seperable.
x, y = make_circles(100, factor=0.1, noise=0.1)
clf = SVC(kernel="linear").fit(x, y)
plt.scatter(x[:, 0], x[:, 1], c=y, s=50, cmap="spring")
plot_svc_decision_function(clf)
plt.show()
# Clearly no linear separation is going to work on this data.
# One way we can adjust to this data is to apply a kernel, which is some
# transformation of the input data. We could use the radial basis function.
r = np.exp(-(x[:, 0] ** 2 + x[:, 1] ** 2))
# if we plot this alongside our data, we can see the effect.
def plot_3D(elev=30, azim=30):
ax = plt.subplot(projection="3d")
ax.scatter3D(x[:, 0], x[:, 1], r, c=y, s=50, cmap="spring")
