train_labels = np.expand_dims(train_labels, 1)
test_labels = np.expand_dims(test_labels, 1)
train_data = np.column_stack((train_samples, train_labels))
test_data = np.column_stack((test_samples, test_labels))
if t_label is not None:
# Redefine the label of the test sample
test_labels[t] = t_label
test_data[t][-1] = t_label
if t_features is not None:
# Redefine the feature values of the test sample
test_samples[t] = t_features
test_data[t] = np.concatenate((t_features, test_labels[t]))
train_set = DataSet(train_samples, train_labels)
test_set = DataSet(test_samples, test_labels)
validation_set = None
data_sets = base.Datasets(train=train_set, test=test_set, validation=validation_set)
# Plot train samples
plot_samples(train_samples, train_labels, plot_pdf=pdf,
plot_title='Train Dataset')
# Plot train samples with test sample
plot_samples(train_samples, train_labels, plot_pdf=pdf,
plot_title='Train Dataset', ref_ex=test_samples[t])
# Plot test samples
plot_samples(test_samples, test_labels, plot_pdf=pdf,
plot_title='Test Dataset',
val_labels = train_labels[:n_val]
train_labels = train_labels[n_val:]
if t_label is not None:
# Redefine the label of the test sample
test_labels[t] = t_label
if t_features is not None:
# Redefine the feature values of the test sample
test_samples[t] = t_features
# Adjust the shape
train_labels = np.expand_dims(train_labels, 1)
test_labels = np.expand_dims(test_labels, 1)
val_labels = np.expand_dims(val_labels, 1)
train_set = DataSet(train_samples, train_labels)
test_set = DataSet(test_samples, test_labels)
validation_set = DataSet(val_samples, val_labels)
data_sets = base.Datasets(train=train_set,
test=test_set,
validation=validation_set)
# Plot train samples
plot_samples(train_samples, train_labels, plot_title='Train Dataset')
# Plot train samples with test sample
plot_samples(train_samples,
train_labels,
plot_title='Train Dataset',
ref_ex=test_samples[t])
from src.InterpretableSpn import InterpretableSpn
from src.influence.dataset import DataSet # for train and test set creation
from src.help_functions import *
from prettytable import PrettyTable
# Get train and test set
num_train_samples = 5000
num_test_samples = 10000
(train_images,
train_labels), (test_images,
test_labels) = load_mnist(num_train_samples,
num_test_samples,
normalization=False)
train_set = DataSet(train_images, np.expand_dims(train_labels, 1))
test_set = DataSet(test_images, np.expand_dims(test_labels, 1))
validation_set = None
data_sets = base.Datasets(train=train_set,
test=test_set,
validation=validation_set)
label_idx = 784
num_classes = 10
batch_size = 1
output_path = "/home/ml-mrothermel/projects/Interpreting-SPNs/output/spns"
file_name = "tf_mnist_spn_9"
# Import a trained, saved and converted model with new placeholders
sample_placeholder = tf.placeholder(dtype=np.float32,