converter = RCNLPWordVectorConverter(resize=-1, pca_model=pca_model)
# end if
# >> 3. Array for results
average_success_rate = np.array([])
# For each samples
for s in range(0, args.samples):
print("#")
# >> 5. Prepare training and test set.
training_set_indexes = np.arange(0, 100, 1)[s:s+args.training_size]
test_set_indexes = np.delete(np.arange(0, 100, 1), training_set_indexes, axis=0)[:args.test_size]
# >> 6. Create Echo Word Classifier
classifier = RCNLPEchoWordClassifier(size=rc_size, input_scaling=rc_input_scaling, leak_rate=rc_leak_rate,
input_sparsity=rc_input_sparsity, converter=converter, n_classes=2,
spectral_radius=rc_spectral_radius, w_sparsity=rc_w_sparsity, w=w)
# >> 7. Add examples
for author_index, author_id in enumerate((args.author1, args.author2)):
author_path = os.path.join(args.dataset, "total", author_id)
for file_index in training_set_indexes:
classifier.add_example(os.path.join(author_path, str(file_index) + ".txt"), author_index)
# end for
# end for
# >> 8. Train model
classifier.train()
# >> 9. Test model performance
success = 0.0
average_success_rate = np.array([])
# n-Fold cross validation
for k in range(0, args.k):
# Info
print(u"K-Fold {}".format(k))
# Prepare training and test set
test_set_indexes = indexes[k]
training_set_indexes = indexes
training_set_indexes = np.delete(training_set_indexes, k, axis=0)
training_set_indexes.shape = (n_texts - n_fold_samples)
# Create Echo Word Classifier
classifier = RCNLPEchoWordClassifier(size=rc_size, input_scaling=rc_input_scaling, leak_rate=rc_leak_rate,
input_sparsity=rc_input_sparsity, converter=converter, n_classes=2,
spectral_radius=rc_spectral_radius, w_sparsity=rc_w_sparsity)
# Add examples
print(u"Adding examples...")
for training_index in training_set_indexes:
training_text_path = os.path.join(args.dataset, text_codes[training_index] + ".txt")
training_text_author = texts_data[text_codes[training_index]]
classifier.add_example(training_text_path, training_text_author)
# end for
# Train model
print(u"Training...")
classifier.train()
# Test model performance
# >> 4. For each samples
for reservoir_size in reservoir_sizes:
print("Reservoir size %d" % reservoir_size)
# Average success rate for this training size
reservoir_size_average_success_rate = np.array([])
# >> 4. Try n time
for s in range(0, args.samples):
try:
# >> 6. Create Echo Word Classifier
classifier = RCNLPEchoWordClassifier(
size=reservoir_size,
input_scaling=rc_input_scaling,
leak_rate=rc_leak_rate,
input_sparsity=rc_input_sparsity,
converter=converter,
n_classes=args.n_authors,
spectral_radius=rc_spectral_radius,
w_sparsity=rc_w_sparsity)
# >> 7. Add examples
for author_index, author_id in enumerate(
np.arange(1, args.n_authors + 1, 1)):
author_path = os.path.join(args.dataset, "total",
author_id)
for file_index in training_set_indexes:
classifier.add_example(
os.path.join(author_path,
str(file_index) + ".txt"),
author_index)
# W matrix
w = mdp.numx.random.choice([0.0, 1.0], (rc_size, rc_size), p=[1.0 - rc_w_sparsity, rc_w_sparsity])
w[w == 1] = mdp.numx.random.rand(len(w[w == 1]))
# >> 4. For each leak rate
for index, leaky_rate in enumerate((args.lr1, args.lr2)):
print("Leaky rate %f" % leaky_rate)
# >> 5. Prepare training and test set.
training_set_indexes = np.arange(1, 100, 1)
test_set_indexes = np.array([0])
# >> 6. Create Echo Word Classifier
classifier = RCNLPEchoWordClassifier(size=rc_size, input_scaling=rc_input_scaling, leak_rate=leaky_rate,
input_sparsity=rc_input_sparsity, converter=converter, n_classes=2,
spectral_radius=rc_spectral_radius, w_sparsity=rc_w_sparsity, w=w)
# >> 7. Add examples
for author_index, author_id in enumerate((args.author1, args.author2)):
author_path = os.path.join(args.dataset, "total", author_id)
for file_index in training_set_indexes:
classifier.add_example(os.path.join(author_path, str(file_index) + ".txt"), author_index)
# end for
# end for
# >> 8. Train model
classifier.train()
# >> 9. Test model performance
author_index = 0
# >> 4. Try n time
for s in range(args.samples):
n_token = 0
# Choose a random author
the_author = np.random.choice(np.arange(1, 51, 1))
# >> 5. Prepare training and test set.
training_set_indexes = np.random.choice(np.arange(0, 100), size=1, replace=False)
test_set_indexes = np.delete(np.arange(0, 100, 1), training_set_indexes)[:args.test_size]
negatives_set_indexes = np.arange(0, args.negatives, 1)
other_authors = np.delete(np.arange(1, 51, 1), the_author-1)
# >> 6. Create Echo Word Classifier
classifier = RCNLPEchoWordClassifier(size=rc_size, input_scaling=rc_input_scaling, leak_rate=rc_leak_rate,
input_sparsity=rc_input_sparsity, converter=converter, n_classes=2,
spectral_radius=rc_spectral_radius, w_sparsity=rc_w_sparsity, w=w)
# >> 7. Add authors examples
author_path = os.path.join(args.dataset, "total", str(the_author))
for file_index in training_set_indexes:
file_path = os.path.join(author_path, str(file_index) + ".txt")
classifier.add_example(file_path, 0)
n_token += get_n_token(file_path)
# end for
# >> 8. Add negative examples
others_path = os.path.join(args.dataset, "total", "others")
for file_index in negatives_set_indexes:
file_path = os.path.join(others_path, str(file_index) + ".txt")
classifier.add_example(file_path, 1)
