def create_dataset(self, opt, means_x, std_x, name):
means_x = list(means_x)
variances_x = [np.eye(2) * std_x for _ in means_x]
priors_x = [1.0 / len(means_x) for _ in means_x]
dataset_x = sample_GMM(opt.dataset_size_x,
means_x,
variances_x,
priors_x,
sources=('features', ))
save_path_x = opt.result_dir + name
# plot_GMM(dataset, save_path)
## reconstruced x
X_dataset = dataset_x.data['samples']
X_targets = dataset_x.data['label']
self.set_means(means_x)
self.set_priors(priors_x)
self.set_variances(variances_x)
self.set_dataset(X_dataset)
fig_mx, ax = plt.subplots(nrows=1, ncols=1, figsize=(4.5, 4.5))
ax.scatter(X_dataset[:, 0],
X_dataset[:, 1],
c=cm.Set1(X_targets.astype(float) / opt.input_dim / 2.0),
edgecolor='none',
alpha=0.5)
ax.set_xlim(-3, 3)
ax.set_ylim(-3.5, 3.5)
ax.set_xlabel('$x_1$')
ax.set_ylabel('$x_2$')
ax.axis('on')
plt.savefig(save_path_x, transparent=True, bbox_inches='tight')
def create_dataset_test(self, opt, std_x, name):
variances_x = self.get_variances()
priors_x = self.get_priors()
means_x = self.get_means()
# create X dataset
datasetX_test = sample_GMM(opt.dataset_size_x_test,
means_x,
variances_x,
priors_x,
sources=('features', ))
save_path = opt.result_dir + name
# plot_GMM(dataset, save_path)
## reconstruced x
X_np_data_test = datasetX_test.data['samples']
X_targets_test = datasetX_test.data['label']
self.set_dataset_test(X_np_data_test)
self.set_label_test(X_targets_test)
fig_mx, ax = plt.subplots(nrows=1, ncols=1, figsize=(4.5, 4.5))
ax.scatter(X_np_data_test[:, 0],
X_np_data_test[:, 1],
c=cm.Set1(
X_targets_test.astype(float) / opt.input_dim / 2.0),
edgecolor='none',
alpha=0.5)
ax.set_xlim(-3, 3)
ax.set_ylim(-3.5, 3.5)
ax.set_xlabel('$x_1$')
ax.set_ylabel('$x_2$')
ax.axis('on')
plt.savefig(save_path, transparent=True, bbox_inches='tight')
# means = map(lambda x: np.array(x), [[0, 0]])
means_x = map(lambda x: np.array(x), [[0, 0],
[2, 2],
[-2, -2],
[2, -2],
[-2, 2]])
means_x = list(means_x)
std_x = 0.04
variances_x = [np.eye(2) * std_x for _ in means_x]
priors_x = [1.0/len(means_x) for _ in means_x]
gaussian_mixture = GMM_distribution(means=means_x,
variances=variances_x,
priors=priors_x)
dataset_x = sample_GMM(dataset_size_x, means_x, variances_x, priors_x, sources=('features', ))
save_path_x = result_dir + 'X_gmm_data_train.pdf'
# plot_GMM(dataset, save_path)
## reconstruced x
X_dataset = dataset_x.data['samples']
X_targets = dataset_x.data['label']
fig_mx, ax = plt.subplots(nrows=1, ncols=1, figsize=(4.5, 4.5))
ax.scatter(X_dataset[:, 0], X_dataset[:, 1], c=cm.Set1(X_targets.astype(float)/input_dim/2.0),
edgecolor='none', alpha=0.5)
ax.set_xlim(-3, 3); ax.set_ylim(-3.5, 3.5)
ax.set_xlabel('$x_1$'); ax.set_ylabel('$x_2$')
ax.axis('on')
plt.savefig(save_path_x, transparent=True, bbox_inches='tight')
""" Create dataset """
# create X dataset
means_x = map(lambda x: np.array(x), [[0, 0],
[2, 2],
[-2, -2],
[2, -2],
[-2, 2]])
means_x = list(means_x)
std_x = 0.04
variances_x = [np.eye(2) * std_x for _ in means_x]
priors_x = [1.0/len(means_x) for _ in means_x]
dataset_x = sample_GMM(dataset_size_x, means_x, variances_x, priors_x, sources=('features', ))
save_path_x = result_dir + 'X_train.png'
# plot x
X_dataset = dataset_x.data['samples']
X_targets = dataset_x.data['label']
fig_mx, ax = plt.subplots(nrows=1, ncols=1, figsize=(4.5, 4.5))
ax.scatter(X_dataset[:, 0], X_dataset[:, 1], c=cm.Set1(X_targets.astype(float)/x_dim/2.0),
edgecolor='none', alpha=0.5)
ax.set_xlim(-3, 3); ax.set_ylim(-3.5, 3.5)
ax.set_xlabel('$x$ train')
ax.axis('on')
plt.savefig(save_path_x, transparent=True, bbox_inches='tight')
# create Y dataset