def shuffle_test(gml_file, auth_time, N_getter, tempfile, _iter, window,
_small):
"""
The procedure generate a file with estimated coefficients
for instances of randomized dataset.
Keyword arguments:
gml_file --- the graph in .gml format
auth_time --- list of activation times for authority nodes
N_getter --- function object that calculates the values of A(x, y)
and N(x,y). Can be calc.getNs or calc.getNs2.
tempfile --- name of the file the coefficients will be writen to
_iter --- number of iterations - runs of the randomization and estimation
steps
window --- length of the window for one time step
_small -- if True, the estimation will be run on restricted, only
with nodes that eventually become active
"""
print "...reading graph..."
g = read(gml_file)
nodes = dict()
unactive_ids = []
for node in g.vs:
if int(node["regtime"]) == -1:
unactive_ids.append(node["id"])
else:
nodes[node["id"]] = node
f = open(tempfile, 'a')
(A, N) = N_getter(unactive_ids, nodes, auth_time, window, False, _small)
f.write(" ".join(str(num) for num in logreg(A, N)) + "\n")
all_time = [node["regtime"]
for node in g.vs if node["regtime"] != -1] + auth_time
for j in range(0, _iter):
print j
all_time = permutation(all_time)
auth_time = []
i = 0
for _id, node in nodes.iteritems():
while len(auth_time) != 11 and all_time[i] != -1:
auth_time.append(all_time[i])
i = i + 1
if node["regtime"] == -1:
continue
nodes[_id]["regtime"] = all_time[i]
i = i + 1
(A, N) = N_getter(unactive_ids, nodes, auth_time, window, False,
_small)
f.write(" ".join(str(num) for num in logreg(A, N)) + "\n")
f.close()
def run_all():
import pandas as pd
from ds import decision_tree
from knn import knn
from logreg import logreg
from sv import sv
from ensemble_methods import ensemble_methods
from xg import xgb
print('Running All')
acc_ds = decision_tree()
acc_knn = knn()
acc_log = logreg()
acc_xg = xgb()
acc_svc = sv()
acc_rf, acc_ab, acc_gb = ensemble_methods()
# Model Performance
models = pd.DataFrame({
'Model': [
'XGBoost', 'Logistic Regression', 'KNN', 'Support Vector Machines',
'Gradient Boosting', 'Random Forest', 'Decision Tree', 'ADABoost'
],
'Score':
[acc_xg, acc_log, acc_knn, acc_svc, acc_gb, acc_rf, acc_ds, acc_ab]
})
models = models.sort_values(by='Score', ascending=True)
print models.sort_values(by='Score', ascending=False)
def _train(self, X, Y):
phi = makePhi(X,self.M)
n,m = phi.shape
self.result = logreg.logreg(phi, Y, self.params['lamduh'], opt=scipy.optimize.fmin_bfgs, printInfo=self.printInfo)
w = self.result[0][:m]
b = self.result[0][m]
return w,b
def _train(self, X, Y):
phi = makePhi(X, self.M)
n, m = phi.shape
self.result = logreg.logreg(phi,
Y,
self.params['lamduh'],
opt=scipy.optimize.fmin_bfgs,
printInfo=self.printInfo)
w = self.result[0][:m]
b = self.result[0][m]
return w, b
def train(lamduh=0.1, basisfunc='lin', plot=False, optimizePrint=False, name='ls'):
# handle basis function
if basisfunc=='lin': M=1
elif basisfunc=='quad': M=2
else: raise Exception('Value "%s" for basisfunc must be either "lin" or "quad"'%(basisfunc))
# parameters
print '======Training======'
print 'lambda = ' + str(lamduh)
# load data from csv files
train = loadtxt('data/data_'+name+'_train.csv')
X = train[:,0:2]
Y = train[:,2:3]
# Carry out training.
phi = makePhi(X,M)
n,m = phi.shape
result = logreg.logreg(phi, Y, lamduh, opt=scipy.optimize.fmin_bfgs, printInfo=optimizePrint)
w = result[0][:m]
b = result[0][m]
# Define the predictLR(x) function, which uses trained parameters
predictLR = makePredictor(w,b,M,mode='logreg')
# get training error
tErr = getError(X, Y, w, b, mode='logreg')
# plot training results
if plot:
plotDecisionBoundary(X, Y, predictLR, [0.5], title = 'LR Train with ${\lambda}=%s$' %(lamduh))
print '======Validation======'
# load data from csv files
validate = loadtxt('data/data_'+name+'_validate.csv')
X = validate[:,0:2]
Y = validate[:,2:3]
# plot validation results
if plot:
plotDecisionBoundary(X, Y, predictLR, [0.5], title = 'LR Validate with ${\lambda}=%s$' %(lamduh))
# get validation error
vErr = getError(X, Y, w, b, mode='logreg')
## return training and validation error
return numpy.array([tErr, vErr])
def main():
t0 = time()
args = get_max_norm_args()
# labels = np.loadtxt("../data/labels", delimiter="\n", dtype=np.str)[:: 2]
# labels = {ii: ll[:-2] for ii, ll in enumerate(labels)}
X, y = get_data('../data/pure_landmarks_gender.npy')
scores = []
for k in reversed(range(1, len(args) + 1)):
print(k, flush=True)
# scores.append(logreg(filter_points(X, args[:k]), y))
scores.append(logreg(filter_with_pca(X, k), y))
np.save('log_reg_scores_pca', np.array(scores))
print(time() - t0)
# training_data, validation_data, test_data = cPickle.load(f)
# hiddenunits = [200, 300, 400, 500, 600]
# learning_rates = [0.01, 0.03, 0.05, 0.07, 0.09, 0.1]
# iterations = [1000, 2000, 5000, 10000, 20000 ]
# f.close()
usps_test_data = [[],[]]
for j in range(10):
for image in glob.glob("Numerals/" + str(j) + "/*.png"):
img = Image.open(image)
img = img.resize((new_height, new_width))
pix = numpy.array(img.getdata(), dtype = 'float64')
pix = [1]*784 - (pix/([255]*784))
usps_test_data[0].append(pix)
usps_test_data[1].append(j)
# print numpy.array(usps_test_data[0][0])
# print training_data[0][0]
# print numpy.array(usps_test_data[1])
print "LOGISTIC REGRESSION"
logreg.logreg(training_data, validation_data, test_data, usps_test_data, 10, 0.0005)
print "SINGLE LAYER NEURAL NETWORKS"
neuralnetwork.neuralnetwork(training_data, validation_data, test_data, usps_test_data)
print "CONVOLUTIONAL NEURAL NETWORKS"
tensor.tensorflow(usps_test_data[0], usps_test_data[1])
def train(lamduh=0.1,
basisfunc='lin',
plot=False,
optimizePrint=False,
name='ls'):
# handle basis function
if basisfunc == 'lin': M = 1
elif basisfunc == 'quad': M = 2
else:
raise Exception(
'Value "%s" for basisfunc must be either "lin" or "quad"' %
(basisfunc))
# parameters
print '======Training======'
print 'lambda = ' + str(lamduh)
# load data from csv files
train = loadtxt('data/data_' + name + '_train.csv')
X = train[:, 0:2]
Y = train[:, 2:3]
# Carry out training.
phi = makePhi(X, M)
n, m = phi.shape
result = logreg.logreg(phi,
Y,
lamduh,
opt=scipy.optimize.fmin_bfgs,
printInfo=optimizePrint)
w = result[0][:m]
b = result[0][m]
# Define the predictLR(x) function, which uses trained parameters
predictLR = makePredictor(w, b, M, mode='logreg')
# get training error
tErr = getError(X, Y, w, b, mode='logreg')
# plot training results
if plot:
plotDecisionBoundary(X,
Y,
predictLR, [0.5],
title='LR Train with ${\lambda}=%s$' % (lamduh))
print '======Validation======'
# load data from csv files
validate = loadtxt('data/data_' + name + '_validate.csv')
X = validate[:, 0:2]
Y = validate[:, 2:3]
# plot validation results
if plot:
plotDecisionBoundary(X,
Y,
predictLR, [0.5],
title='LR Validate with ${\lambda}=%s$' %
(lamduh))
# get validation error
vErr = getError(X, Y, w, b, mode='logreg')
## return training and validation error
return numpy.array([tErr, vErr])
def classification_start(self):
try:
self.Hitung.configure(state='disabled')
self.MulaiKlasifikasi.configure(state='disabled')
self.Scrolledtext1.configure(state='normal')
file = io.StringIO()
with redirect_stdout(file):
if v2.get() == 3:
if v1.get() == 1:
obj_LR = logreg(bow_tfidf_train, train, bow_tfidf_test,
test)
roc_name = "ROC BagOfWords Logistic Regression"
csv_name = "BagOfWords Logistic Regression"
elif v1.get() == 2:
obj_LR = logreg(d2v_train, train, d2v_test, test)
roc_name = "ROC Doc2Vec Logistic Regression"
csv_name = "Doc2Vec Logistic Regression"
pred_res = obj_LR.logreg()
elif v2.get() == 2:
if v1.get() == 1:
obj_SVM = SVM(bow_tfidf_train, train, bow_tfidf_test,
test)
roc_name = "ROC BagOfWords Support Vector Machine"
csv_name = "BagOfWords Support Vector Machine"
elif v1.get() == 2:
obj_SVM = SVM(d2v_train, train, d2v_test, test)
roc_name = "ROC Doc2Vec Support Vector Machine"
csv_name = "Doc2Vec Support Vector Machine"
pred_res = obj_SVM.svm()
elif v2.get() == 1:
if v1.get() == 1:
obj_MNB = MNB(bow_tfidf_train, train, bow_tfidf_test,
test)
roc_name = "ROC BagOfWords Multinomial Naive Bayes"
csv_name = "BagOfWords Multinomial Naive Bayes"
elif v1.get() == 2:
obj_MNB = MNB(d2v_train, train, d2v_test, test)
roc_name = "ROC Doc2Vec Multinomial Naive Bayes"
csv_name = "Doc2Vec Multinomial Naive Bayes"
pred_res = obj_MNB.mnb()
obj_csv = save_to_csv(test)
obj_csv.save(pred_res, csv_name)
output = file.getvalue()
self.Scrolledtext1.insert(END, output)
self.Scrolledtext1.configure(state='disabled')
self.Hitung.configure(
state='normal', command=lambda: self.plot(pred_res, roc_name))
self.MulaiKlasifikasi.configure(state='normal')
except NameError:
self.MulaiKlasifikasi.configure(state='normal')
self.Scrolledtext1.configure(state='normal')
self.Scrolledtext1.insert(
END, "Lakukan pembentukan VSM terlebih dahulu!\n")
self.Scrolledtext1.configure(state='disabled')
except ValueError:
self.MulaiKlasifikasi.configure(state='normal')
self.Scrolledtext1.configure(state='normal')
self.Scrolledtext1.insert(
END,
"Masukkan nilai learning rate dan jumlah iterasi. Bilangan yang diinput hanya berupa angka!\n"
)
self.Scrolledtext1.configure(state='disabled')
except AssertionError:
self.MulaiKlasifikasi.configure(state='normal')
self.Scrolledtext1.configure(state='normal')
self.Scrolledtext1.insert(
END,
"Range yang diterima: 0 < learning rate <= 1 dan jumlah iterasi > 0!\n"
)
self.Scrolledtext1.configure(state='disabled')
except AttributeError:
self.MulaiKlasifikasi.configure(state='normal')
self.Scrolledtext1.configure(state='normal')
self.Scrolledtext1.insert(END,
"Fitur tidak bisa bernilai negatif!\n")
self.Scrolledtext1.configure(state='disabled')
