def run_fa(dataset, min_components, max_components):
X, y = load_dataset(dataset)
data = X
n_samples, n_features = data.shape
n_labels = len(np.unique(y))
labels = y
results = []
for n_components in range(min_components, max_components):
print('n_components: ', n_components)
for svd_method in ['lapack', 'randomized']:
scores = []
data = X.copy()
fa = FactorAnalysis(n_components=n_components,
svd_method=svd_method,
random_state=random_state)
t0 = time()
fa.fit(X)
scores.append(n_components)
scores.append(svd_method)
scores.append(time() - t0)
scores.append(fa.score(X))
results.append(scores)
# N-Components vs Log Likelihood
plot_results(np.array(results),
trends_index=1,
x_axis_index=0,
x_axis_label='K-Components',
y_axis_index=[3],
y_axis_label='Log Liklihood',
title=dataset.title() + ': FactorAnalysis',
filename='-'.join(['fa', dataset, 'loglike']))
# N-Components vs Time
plot_results(np.array(results),
trends_index=1,
x_axis_index=0,
x_axis_label='K-Components',
y_axis_index=[2],
y_axis_label='Time',
title=dataset.title() + ': FactorAnalysis',
filename='-'.join(['fa', dataset, 'time']))
results = np.array(results)
np.savetxt('output-csv/' + ('-'.join([dataset, 'fa.csv'])),
results,
delimiter=",",
fmt="%s")
def run_pca(dataset, min_components, max_components):
X, y = load_dataset(dataset)
data = X
n_samples, n_features = data.shape
n_labels = len(np.unique(y))
labels = y
results = []
for n_components in range(min_components, max_components):
print('n_components: ', n_components)
for svd_solver in ['auto', 'full', 'randomized']:
scores = []
data = X.copy()
pca = decomposition.PCA(n_components=n_components, svd_solver=svd_solver, random_state=random_state)
t0 = time()
pca.fit(X)
scores.append(n_components)
scores.append(svd_solver)
scores.append(time() - t0)
scores.append(pca.score(data))
scores.append(pca.explained_variance_ratio_[n_components - 1])
scores.append(np.cumsum(np.round(pca.explained_variance_ratio_, decimals=3) * 100)[n_components - 1])
results.append(scores)
# N-Components vs Score
plot_results(np.array(results), trends_index=1, x_axis_index=0, x_axis_label='K-Components', y_axis_index=[3], y_axis_label='Score', title=dataset.title() + ': PCA', filename='-' . join(['pca', dataset, 'score']))
# N-Components vs Variance
plot_results(np.array(results), trends_index=1, x_axis_index=0, x_axis_label='K-Components', y_axis_index=[4], y_axis_label='Variance Ratio', title=dataset.title() + ': PCA', filename='-' . join(['pca', dataset, 'variance']))
# N-Components vs Variance %
plot_results(np.array(results), trends_index=1, x_axis_index=0, x_axis_label='K-Components', y_axis_index=[5], y_axis_label='% Variance', title=dataset.title() + ': PCA', filename='-' . join(['pca', dataset, 'pvar']))
# N-Components vs Time
plot_results(np.array(results), trends_index=1, x_axis_index=0, x_axis_label='K-Components', y_axis_index=[2], y_axis_label='Time', title=dataset.title() + ': PCA', filename='-' . join(['pca', dataset, 'time']))
results = np.array(results)
np.savetxt('output-csv/' + ('-' . join([dataset, 'pca.csv'])), results, delimiter=",", fmt="%s")
def run_ica(dataset, min_components, max_components):
X, y = load_dataset(dataset)
data = X
n_samples, n_features = data.shape
n_labels = len(np.unique(y))
labels = y
results = []
for n_components in range(min_components, max_components):
print('n_components: ', n_components)
data = X.copy()
ica = decomposition.FastICA(n_components=n_components,
random_state=random_state,
whiten=True)
t0 = time()
ica.fit(X)
elapsed = time() - t0
kurtoses = np.sort(kurtosis(ica.components_))
kurt_vsgauss_vals = []
for component_index in range(min_components, max_components):
scores = []
scores.append(n_components)
scores.append(component_index)
scores.append(time() - t0)
scores.append(ica.n_iter_)
sources = ica.transform(X)
if component_index <= n_components:
kurt = kurtoses[component_index - 1]
kurt_vsgauss = abs(3 - kurt)
else:
kurt = 0
kurt_vsgauss = 0
kurt_vsgauss_vals.append(kurt_vsgauss)
kurt_norm = sum(kurt_vsgauss_vals) / max_components
scores.append(kurt)
scores.append(kurt_vsgauss)
scores.append(kurt_norm)
scores.append(kurtosis(ica.components_, axis=None))
scores.append(kurtosis(sources, axis=None))
results.append(scores)
# N-Components vs Kurtosis (Component)
plot_results(np.array(results),
trends_index=1,
x_axis_index=0,
x_axis_label='Dimensions',
y_axis_index=[4],
y_axis_label='Kurtosis',
title=dataset.title() + ': ICA',
filename='-'.join(['ica', dataset, 'kurt']))
# N-Components vs Kurtosis (vs Gaussian)
plot_results(np.array(results),
trends_index=1,
x_axis_index=0,
x_axis_label='Dimensions',
y_axis_index=[5],
y_axis_label='Kurtosis (vs Gaussian)',
title=dataset.title() + ': ICA',
filename='-'.join(['ica', dataset, 'kurt', 'vsgauss']))
# N-Components vs Kurtosis (vs Gaussian)
plot_results(np.array(results),
constant_index=[1],
constant_value=[max_components - 1],
x_axis_index=0,
x_axis_label='Dimensions',
y_axis_index=[6],
y_axis_label='Kurtosis (%)',
trend_labels=[''],
title=dataset.title() + ': ICA',
filename='-'.join(['ica', dataset, 'kurt', 'norm']))
# N-Components vs Overall Kurtosis (Component)
plot_results(np.array(results),
constant_index=[1],
constant_value=[1],
x_axis_index=0,
x_axis_label='K-Components',
y_axis_index=[7],
y_axis_label='Overall Kurtosis (Component)',
trend_labels=[''],
title=dataset.title() + ': ICA',
filename='-'.join(['ica', dataset, 'kurt', 'overall',
'comp']))
# N-Components vs Overall Kurtosis (Source)
plot_results(np.array(results),
constant_index=[1],
constant_value=[1],
x_axis_index=0,
x_axis_label='K-Components',
y_axis_index=[8],
y_axis_label='Overall Kurtosis (Source)',
trend_labels=[''],
title=dataset.title() + ': ICA',
filename='-'.join(
['ica', dataset, 'kurt', 'overall', 'source']))
# N-Components vs Iter
plot_results(np.array(results),
constant_index=[1],
constant_value=[1],
x_axis_index=0,
x_axis_label='K-Components',
y_axis_index=[3],
y_axis_label='Iterations',
trend_labels=[''],
title=dataset.title() + ': ICA',
filename='-'.join(['ica', dataset, 'iter']))
# N-Components vs Time
plot_results(np.array(results),
constant_index=[1],
constant_value=[1],
x_axis_index=0,
x_axis_label='K-Components',
y_axis_index=[2],
y_axis_label='Time',
trend_labels=[''],
title=dataset.title() + ': ICA',
filename='-'.join(['ica', dataset, 'time']))
results = np.array(results)
np.savetxt('output-csv/' + ('-'.join([dataset, 'ica.csv'])),
results,
delimiter=",",
fmt="%s")
os.system("make all")
os.chdir("../data")
#************************
# RUN INVERSION
os.system("../bin/hv_na")
sys.exit()
#************************
os.chdir("../../")
# Move results to result folder
os.system("mv full_list.txt " + out_folder)
# Save inversion parameters to keep track
os.system("cp NA/data/na.in " + out_folder)
os.system("cp NA/data/hv_files/hv.in " + out_folder)
os.system("cp NA/data/hv_files/NA_MDL/hv_param " + out_folder)
lines = open("NA/data/hv_files/hv.in").readlines()
# Save observed curve to results folder
observed_data = lines[6].split()[0]
os.system("cp NA/data/hv_files/OBS/" + observed_data + " " + out_folder)
# Save model used for litho to results folder
ref_model = lines[8].split()[0]
os.system("cp NA/data/hv_files/REF_MDL/" + ref_model + " " + out_folder)
# Plot results
plot_results(out_folder, observed_data, ref_model)
def optimize_k_means(dataset, min_clusters, max_clusters):
if '-' in dataset:
X, y = load_reduced(dataset)
else:
X, y = load_dataset(dataset)
data = X
n_samples, n_features = data.shape
n_labels = len(np.unique(y))
labels = y
results = []
for n_clusters in range(min_clusters, max_clusters):
print('n_clusters: ', n_clusters)
for init in ['k-means++', 'random']:
scores = []
estimator = KMeans(n_clusters=n_clusters,
init=init,
n_init=10,
random_state=random_state)
t0 = time()
estimator.fit(data)
scores.append(n_clusters)
scores.append(init)
scores.append(time() - t0)
scores.append(estimator.inertia_)
scores.append(metrics.homogeneity_score(labels, estimator.labels_))
scores.append(metrics.completeness_score(labels,
estimator.labels_))
scores.append(metrics.v_measure_score(labels, estimator.labels_))
scores.append(
metrics.adjusted_rand_score(labels, estimator.labels_))
scores.append(
metrics.adjusted_mutual_info_score(labels, estimator.labels_))
scores.append(metrics.silhouette_score(data, estimator.labels_))
results.append(scores)
# N-Clusters vs Inertia per Init Centroid
plot_results(np.array(results),
trends_index=1,
x_axis_index=0,
x_axis_label='K-Clusters',
y_axis_index=[3],
y_axis_label='Inertia',
title=dataset.title() + ': K-means',
filename='-'.join(['km', dataset, 'inertia']))
# N-Clusters vs Homogeneity Score per Init Centroid
plot_results(np.array(results),
trends_index=1,
x_axis_index=0,
x_axis_label='K-Clusters',
y_axis_index=[4],
y_axis_label='Homogeneity Score',
title=dataset.title() + ': K-means',
filename='-'.join(['km', dataset, 'homogeneity']))
# N-Clusters vs Completeness Score per Init Centroid
plot_results(np.array(results),
trends_index=1,
x_axis_index=0,
x_axis_label='K-Clusters',
y_axis_index=[5],
y_axis_label='Completeness Score',
title=dataset.title() + ': K-means',
filename='-'.join(['km', dataset, 'completeness']))
# N-Clusters vs V-Measure Score per Init Centroid
plot_results(np.array(results),
trends_index=1,
x_axis_index=0,
x_axis_label='K-Clusters',
y_axis_index=[6],
y_axis_label='V-Measure Score',
title=dataset.title() + ': K-means',
filename='-'.join(['km', dataset, 'vmeasure']))
# N-Clusters vs Adjusted Random Score per Init Centroid
plot_results(np.array(results),
trends_index=1,
x_axis_index=0,
x_axis_label='K-Clusters',
y_axis_index=[7],
y_axis_label='Adjusted Random Score',
title=dataset.title() + ': K-means',
filename='-'.join(['km', dataset, 'adjustedrand']))
# N-Clusters vs Adjusted Mutual Information Score per Init Centroid
plot_results(np.array(results),
trends_index=1,
x_axis_index=0,
x_axis_label='K-Clusters',
y_axis_index=[8],
y_axis_label='Adjusted Mutual Information Score',
title=dataset.title() + ': K-means',
filename='-'.join(['km', dataset, 'adjustedmi']))
# N-Clusters vs Silhouette Score per Init Centroid
plot_results(np.array(results),
trends_index=1,
x_axis_index=0,
x_axis_label='K-Clusters',
y_axis_index=[9],
y_axis_label='Silhouette Score',
title=dataset.title() + ': K-means',
filename='-'.join(['km', dataset, 'silhouette']))
# N-Clusters vs Time per Init Centroid
plot_results(np.array(results),
trends_index=1,
x_axis_index=0,
x_axis_label='K-Clusters',
y_axis_index=[2],
y_axis_label='Time',
title=dataset.title() + ': K-means',
filename='-'.join(['km', dataset, 'time']))
# N-Clusters vs V-Measures per Scoring Method w/o CV
plot_results(np.array(results),
constant_index=[1],
constant_value=['k-means++'],
x_axis_index=0,
x_axis_label='K-Clusters',
y_axis_index=[4, 5, 6],
y_axis_label='Score',
trend_labels=['Homogeneity', 'Completeness', 'V-Measure'],
title=dataset.title() + ': K-means',
filename='-'.join(['km', dataset, 'kmeans', 'vmeasure']))
# N-Clusters vs Score per Scoring Method w/o CV
plot_results(np.array(results),
constant_index=[1],
constant_value=['k-means++'],
x_axis_index=0,
x_axis_label='K-Clusters',
y_axis_index=[6, 8, 9],
y_axis_label='Score',
trend_labels=[
'V-Measure', 'Adjusted Mutual Information', 'Silhouette'
],
title=dataset.title() + ': K-means',
filename='-'.join(['km', dataset, 'kmeans', 'scores']))
results = np.array(results)
np.savetxt('output-csv/' + ('-'.join([dataset, 'kmeans.csv'])),
results,
delimiter=",",
fmt="%s")
def run_rp(dataset, min_components, max_components):
X, y = load_dataset(dataset)
data = X
n_samples, n_features = data.shape
n_labels = len(np.unique(y))
labels = y
results = []
for n_components in range(min_components, max_components):
print('n_components: ', n_components)
for max_iters in [10, 40, 100, 500]:
scores = []
times = []
components = []
kurtoses = []
losses = []
for iters in range(0, max_iters):
data = X.copy()
rp = GaussianRandomProjection(n_components=n_components)
t0 = time()
rp.fit(X)
times.append(time() - t0)
components.append(rp.n_components_)
kurtoses.append(kurtosis(rp.components_, axis=None))
matrix = rp.components_
new_data = rp.transform(data)
new_data_inv = np.dot(new_data, matrix)
loss = metrics.mean_squared_error(data, new_data_inv)
losses.append(loss)
scores.append(n_components)
scores.append(max_iters)
scores.append(np.mean(np.array(times)))
scores.append(np.mean(np.array(components)))
scores.append(np.mean(np.array(kurtoses)))
scores.append(np.std(np.array(kurtoses)))
scores.append(np.mean(np.array(losses)))
scores.append(np.std(np.array(losses)))
results.append(scores)
# N-Components vs Loss
plot_results(np.array(results),
trends_index=1,
x_axis_index=0,
x_axis_label='K-Components',
y_axis_index=[6],
y_axis_label='Reconstruction Error',
title=dataset.title() + ': RP',
filename='-'.join(['rp', dataset, 'loss']))
# N-Components vs Loss (STD)
plot_results(np.array(results),
trends_index=1,
x_axis_index=0,
x_axis_label='K-Components',
y_axis_index=[7],
y_axis_label='Reconstruction Error (STD)',
title=dataset.title() + ': RP',
filename='-'.join(['rp', dataset, 'lossstd']))
# N-Components vs Kurtosis
plot_results(np.array(results),
trends_index=1,
x_axis_index=0,
x_axis_label='K-Components',
y_axis_index=[4],
y_axis_label='Kurtosis',
title=dataset.title() + ': RP',
filename='-'.join(['rp', dataset, 'kurtosis']))
# N-Components vs Kurtosis (STD)
plot_results(np.array(results),
trends_index=1,
x_axis_index=0,
x_axis_label='K-Components',
y_axis_index=[5],
y_axis_label='Kurtosis (STD)',
title=dataset.title() + ': RP',
filename='-'.join(['rp', dataset, 'kurtstd']))
# N-Components vs Components
plot_results(np.array(results),
trends_index=1,
x_axis_index=0,
x_axis_label='K-Components',
y_axis_index=[3],
y_axis_label='Components',
title=dataset.title() + ': RP',
filename='-'.join(['rp', dataset, 'comp']))
# N-Components vs Time
plot_results(np.array(results),
trends_index=1,
x_axis_index=0,
x_axis_label='K-Components',
y_axis_index=[2],
y_axis_label='Time',
title=dataset.title() + ': RP',
filename='-'.join(['rp', dataset, 'time']))
results = np.array(results)
np.savetxt('output-csv/' + ('-'.join([dataset, 'rp.csv'])),
results,
delimiter=",",
fmt="%s")
def main():
# Get arguments
ny_data = args.ny_data
mu_style = args.mu_style
dist_style = args.dist_style
numrunrange = []
#### Set experimental settings in paper
pi1 = 0.4
hyp_style = 1
top_arms = 10
alpha0 = 0.1 # generally 0.1
FDRrange = [0]
algrange = [0, 1] # both 0: best-arm MAB and 1: Uniform sampling
FDR = 0
# Power or FDR plots
if args.power_plot == 1:
punif = 0
else:
punif = 1
if (dist_style == 1):
mu_gap = 0.2
mu_best = 8
num_hyp = 500
# fix_na = [50]
# fix_tt = [300]
# armrange = np.arange(10, 121, 10)
# truncrange = np.arange(100, 801, 100)
armrange = np.arange(50, 51, 10)
truncrange = np.arange(100, 701, 100)
fix_na = [30]
fix_tt = [300]
if punif == 1:
armrange = [30]
algrange = [0]
truncrange = [200]
pi1range = np.arange(0.1, 1, 0.1)
FDRrange = [0, 3, 5]
num_hyp = 1000
fix_pi1 = [0.4]
plot_numrun = 80
plot_start = 1
elif (dist_style == 0):
mu_gap = 0.3
mu_best = 0.7
num_hyp = 50
# fix_na = [50]
# fix_tt = [5000]
# armrange = np.arange(5, 36, 5)
# truncrange = np.arange(5000, 25001, 1000)
truncrange = np.arange(5000, 10001, 5000)
armrange = np.arange(5, 11, 5)
fix_na = [10]
fix_tt = [10000]
# Overwrite all settings if New Yorker data
if (ny_data == 1):
dist_style = 0
mu_gap = 0
mu_style = 0
mu_best = 0
num_hyp = 30
top_arms = 5
fix_tt = [13000]
# armrange = np.arange(10, 81, 10)
truncrange = np.arange(13000, 13001, 1000)
armrange = np.arange(10, 21, 10)
##### Plot results
# Plot vs. FDR
if args.power_plot == 0:
# Plot over pi1
plot_results_punif(truncrange,
armrange, [0],
dist_style,
mu_gap,
mu_style,
hyp_style,
pi1range,
num_hyp,
0,
0,
top_arms,
FDRrange,
mu_best,
alpha0,
plot_numrun=plot_numrun,
punif=1,
plot_start=plot_start)
# Plot over time
plot_results_punif(truncrange,
armrange,
algrange,
dist_style,
mu_gap,
mu_style,
hyp_style,
fix_pi1,
num_hyp,
0,
0,
top_arms, [0],
mu_best,
alpha0,
plot_numrun=plot_numrun,
punif=1,
plot_start=plot_start)
# Plot vs. TT
if (ny_data == 0):
plot_results(truncrange, fix_na, algrange, dist_style, mu_gap,
mu_style, hyp_style, pi1, num_hyp, 0, 0, top_arms, FDR,
mu_best, 0, 0, 0)
# Plot vs. number of arms
plot_results(fix_tt, armrange, algrange, dist_style, mu_gap, mu_style,
hyp_style, pi1, num_hyp, 0, 0, top_arms, FDR, mu_best, 0, 0,
0)
maximum_flow,
lower_bound_ramping,
upper_bound_ramping,
origin_node_line_relationship,
end_node_line_relationship)
#model.pprint()
solver = SolverFactory("cplex")
results = solver.solve(model,
tee = True)
results_file = open('results.csv', 'w')
results_file.write("generating unit, time_period, generation_unit_power, flow" + "\n")
for generating_unit in model.indexes_generating_units:
for time_period in model.indexes_time_periods:
results_file.write(str(generating_unit) + "," + str(time_period) + "," + str(model.power_generating_units[generating_unit, time_period].value) + "," + str(model.flow[generating_unit, time_period].value) + "\n")
plot_results(model)
from calculate_multicollinearity import * from train_models import * from evaluate_models_performance import * from evaluate_feature_importance import * from create_table_for_performance_results import * from plot_results import * calculate_multicollinearity() train_models() evaluate_models_performance() evaluate_feature_importance() create_table_for_performance_results() plot_results()
def optimize_em(dataset, min_components, max_components):
if '-' in dataset:
X, y = load_reduced(dataset)
else:
X, y = load_dataset(dataset)
data = X
n_samples, n_features = data.shape
n_labels = len(np.unique(y))
labels = y
results = []
for n_components in range(min_components, max_components):
print('n_components: ', n_components)
for init_params in ['kmeans', 'random']:
for covariance_type in ['full', 'tied', 'diag', 'spherical']:
scores = []
estimator = GaussianMixture(n_components=n_components,
init_params=init_params,
n_init=10,
random_state=random_state,
covariance_type=covariance_type,
reg_covar=1e-2)
t0 = time()
estimator.fit(data)
scores.append(n_components)
scores.append(init_params)
scores.append(covariance_type)
scores.append(time() - t0)
scores.append(estimator.n_iter_)
scores.append(estimator.aic(data))
scores.append(estimator.bic(data))
predictions = estimator.predict(data)
scores.append(metrics.homogeneity_score(labels, predictions))
scores.append(metrics.completeness_score(labels, predictions))
scores.append(metrics.v_measure_score(labels, predictions))
scores.append(metrics.adjusted_rand_score(labels, predictions))
scores.append(
metrics.adjusted_mutual_info_score(labels, predictions))
scores.append(metrics.silhouette_score(data, predictions))
#print estimator.converged_
results.append(scores)
# N-Components vs AIC per Init Centroid
plot_results(np.array(results),
constant_index=[2],
constant_value=['full'],
trends_index=1,
x_axis_index=0,
x_axis_label='K-Components',
y_axis_index=[5],
y_axis_label='Akaike Info Criterion',
title=dataset.title() + ': Expectation-Maximization',
filename='-'.join(['em', dataset, 'akaike']))
# N-Components vs BIC per Init Centroid
plot_results(np.array(results),
constant_index=[2],
constant_value=['full'],
trends_index=1,
x_axis_index=0,
x_axis_label='K-Components',
y_axis_index=[6],
y_axis_label='Bayesian Info Criterion',
title=dataset.title() + ': Expectation-Maximization',
filename='-'.join(['em', dataset, 'bayesian']))
# N-Components vs Homogeneity Score per Init Centroid
plot_results(np.array(results),
constant_index=[2],
constant_value=['full'],
trends_index=1,
x_axis_index=0,
x_axis_label='K-Components',
y_axis_index=[7],
y_axis_label='Homogeneity Score',
title=dataset.title() + ': Expectation-Maximization',
filename='-'.join(['em', dataset, 'homogeneity']))
# N-Components vs Completeness Score per Init Centroid
plot_results(np.array(results),
constant_index=[2],
constant_value=['full'],
trends_index=1,
x_axis_index=0,
x_axis_label='K-Components',
y_axis_index=[8],
y_axis_label='Completeness Score',
title=dataset.title() + ': Expectation-Maximization',
filename='-'.join(['em', dataset, 'completeness']))
# N-Components vs V-Measure Score per Init Centroid
plot_results(np.array(results),
constant_index=[2],
constant_value=['full'],
trends_index=1,
x_axis_index=0,
x_axis_label='K-Components',
y_axis_index=[9],
y_axis_label='V-Measure Score',
title=dataset.title() + ': Expectation-Maximization',
filename='-'.join(['em', dataset, 'vmeasure']))
# N-Components vs Adjusted Random Score per Init Centroid
plot_results(np.array(results),
constant_index=[2],
constant_value=['full'],
trends_index=1,
x_axis_index=0,
x_axis_label='K-Components',
y_axis_index=[10],
y_axis_label='Adjusted Random Score',
title=dataset.title() + ': Expectation-Maximization',
filename='-'.join(['em', dataset, 'adjustedrand']))
# N-Components vs Adjusted Mutual Information Score per Init Centroid
plot_results(np.array(results),
constant_index=[2],
constant_value=['full'],
trends_index=1,
x_axis_index=0,
x_axis_label='K-Components',
y_axis_index=[11],
y_axis_label='Adjusted Mutual Information Score',
title=dataset.title() + ': Expectation-Maximization',
filename='-'.join(['em', dataset, 'adjustedmi']))
# N-Components vs Silhouette Score per Init Centroid
plot_results(np.array(results),
constant_index=[2],
constant_value=['full'],
trends_index=1,
x_axis_index=0,
x_axis_label='K-Components',
y_axis_index=[12],
y_axis_label='Silhouette Score',
title=dataset.title() + ': Expectation-Maximization',
filename='-'.join(['em', dataset, 'silhouette']))
# N-Components vs Time per Init Centroid
plot_results(np.array(results),
constant_index=[2],
constant_value=['full'],
trends_index=1,
x_axis_index=0,
x_axis_label='K-Components',
y_axis_index=[3],
y_axis_label='Time',
title=dataset.title() + ': Expectation-Maximization',
filename='-'.join(['em', dataset, 'time']))
# N-Components vs Criterions per Scoring Method
plot_results(np.array(results),
constant_index=[1, 2],
constant_value=['kmeans', 'full'],
x_axis_index=0,
x_axis_label='K-Components',
y_axis_index=[7, 8, 9],
y_axis_label='Score',
trend_labels=['Homogeneity', 'Completeness', 'V-Measure'],
title=dataset.title() + ': Expectation-Maximization',
filename='-'.join(['em', dataset, 'kmeans', 'vmeasure']))
# N-Components vs V-Measures per Scoring Method
plot_results(np.array(results),
constant_index=[1, 2],
constant_value=['kmeans', 'full'],
x_axis_index=0,
x_axis_label='K-Components',
y_axis_index=[5, 6],
y_axis_label='Score',
trend_labels=['Akaike', 'Bayesian'],
title=dataset.title() + ': Expectation-Maximization',
filename='-'.join(['em', dataset, 'kmeans', 'criterion']))
# N-Components vs Score per Scoring Method
plot_results(np.array(results),
constant_index=[1, 2],
constant_value=['kmeans', 'full'],
x_axis_index=0,
x_axis_label='K-Components',
y_axis_index=[9, 11, 12],
y_axis_label='Score',
trend_labels=[
'V-Measure', 'Adjusted Mutual Information', 'Silhouette'
],
title=dataset.title() + ': Expectation-Maximization',
filename='-'.join(['em', dataset, 'kmeans', 'scores']))
# N-Components vs AIC per Covariance
plot_results(np.array(results),
constant_index=[1],
constant_value=['kmeans'],
trends_index=2,
x_axis_index=0,
x_axis_label='K-Components',
y_axis_index=[5],
y_axis_label='Akaike Info Criterion',
title=dataset.title() + ': Expectation-Maximization',
filename='-'.join(['em', dataset, 'kmeans', 'aic']))
# N-Components vs BIC per Covariance
plot_results(np.array(results),
constant_index=[1],
constant_value=['kmeans'],
trends_index=2,
x_axis_index=0,
x_axis_label='K-Components',
y_axis_index=[6],
y_axis_label='Bayesian Info Criterion',
title=dataset.title() + ': Expectation-Maximization',
filename='-'.join(['em', dataset, 'kmeans', 'bic']))
results = np.array(results)
np.savetxt('output-csv/' + ('-'.join([dataset, 'expmax.csv'])),
results,
delimiter=",",
fmt="%s")
if WB[1]['typ'] == 'D':
ind_wezla = WB[1]['ind']
wart_war_brzeg = WB[1]['wartosc']
iwp = ind_wezla - 1 #indeks wezla w tabeli Pythona
wzmacniacz = 10**14
temp = A[iwp, iwp]
A[iwp, iwp] = temp * wzmacniacz
b[iwp] = temp * wzmacniacz * wart_war_brzeg
if WB[0]['typ'] == 'N':
print('Jeszcze nie zaimplementowano')
if WB[1]['typ'] == 'N':
print('Jeszcze nie zaimplementowano')
print(A)
print('\n')
print(b)
print('\n')
#Obliczenie wyniku
x = np.linalg.solve(A, b)
print(x)
#Wyswietlenie graficznej interpretacji wyniku
plot_results(wezly, x)