def main(args):
config = load_config(args.config_path)
dataset_name = config['dataset_selected']
error_change_threshold = config['error_change_threshold']
batch_size = config['batch_size']
learning_rate = config['optimizer_params']['learning_rate']
epochs = config['nb_epochs']
input_dim = config['input_dimension']
embedding_dimension = config['output_dimension']
n_samples = config['number_of_points']
number_of_test_triplets = config['n_test_triplets']
triplet_multiplier = config['triplets_multiplier']
log_dir = config['log']['path']
hyper_search = config['hyper_search']['activation']
optimizer = config['optimizer']
if hyper_search:
run_hyper_search(config=config)
else:
vec_data, labels = select_dataset(dataset_name,
n_samples=n_samples,
input_dim=input_dim)
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
n = vec_data.shape[0]
logn = int(np.log2(n))
triplet_num = triplet_multiplier * logn * n * embedding_dimension
bs = min(batch_size, triplet_num)
if not os.path.exists(log_dir):
os.makedirs(log_dir)
experiment_name = 'tste_' + \
'data_' + dataset_name + \
'_error_change_threshold_' + str(error_change_threshold) + \
'_input_dim_' + str(input_dim) + \
'_output_dim_' + str(embedding_dimension) + \
'_originaldimension_' + str(vec_data.shape[1]) + \
'_triplet_num_' + str(triplet_multiplier) + \
'_n_pts_' + str(n) + \
'_lr_' + str(learning_rate) + \
'_optimizer_' + str(optimizer) + \
'_bs_' + str(batch_size)
# create a logging file for extensive logging
logging_path = os.path.join(log_dir, experiment_name + '.log')
logger = logging_util.my_custom_logger(logger_name=logging_path,
level=logging.INFO)
logger.info('Name of Experiments: ' + experiment_name)
logger.info('Logging Path:' + logging_path)
logger.info('Dataset Name: ' + dataset_name)
logger.info('Error Change Threshold: ' + str(error_change_threshold))
logger.info('Epochs: ' + str(epochs))
logger.info('Learning Rate: ' + str(learning_rate))
logger.info('Number of Points: ' + str(n))
logger.info('Input Dimension: ' + str(input_dim))
logger.info('Output Dimension: ' + str(embedding_dimension))
logger.info('Number of Test Triplets: ' + str(number_of_test_triplets))
logger.info('Triplet Multiplier: ' + str(triplet_multiplier))
logger.info('Batch Size: ' + str(batch_size))
train_triplets_dataset = TripletBatchesDataset(vec_data, labels,
triplet_num, bs, device)
logger.info('Computing TSTE...')
x, loss_history, triplet_error_history, time_taken, time_history = tste.t_ste_adam(
triplets=train_triplets_dataset.trips_data_indices,
n=n,
emb_dim=embedding_dimension,
epochs=epochs,
batch_size=bs,
learning_rate=learning_rate,
device=device,
logger=logger,
error_change_threshold=error_change_threshold)
logger.info('Evaluating the computed embeddings...')
# compute triplet error for train and test data
train_error = train_triplets_dataset.triplet_error(x)
test_triplets_dataset = TripletBatchesDataset(vec_data, labels,
number_of_test_triplets,
1000, device)
test_error = test_triplets_dataset.triplet_error(x)
procrustes_error = procrustes_disparity(vec_data, x)
knn_error_ord_emb, knn_error_true_emb = knn_classification_error(
x, vec_data, labels)
# sample points for tsne visualization
subsample = np.random.permutation(n)[0:500]
x = x[subsample, :]
sub_labels = labels[subsample]
x_embedded = TSNE(n_components=2, perplexity=15,
learning_rate=10).fit_transform(x)
fig, ax = plt.subplots(1, 1)
ax.scatter(x_embedded[:, 0], x_embedded[:, 1], s=3, c=sub_labels)
fig.savefig(os.path.join(log_dir, experiment_name + '.png'))
logger.info('Name of Experiments: ' + experiment_name)
logger.info('Epochs: ' + str(epochs))
logger.info('Time Taken: ' + str(time_taken) + ' seconds.')
logger.info('Train Error: ' + str(train_error))
logger.info('Test Error: ' + str(test_error))
logger.info('Procrustes Disparity: ' + str(procrustes_error))
logger.info('kNN Classification Error on ground-truth: ' +
str(knn_error_true_emb))
logger.info('kNN Classification Error on embedding: ' +
str(knn_error_ord_emb))
results = {
'train_error': train_error,
'test_error': test_error,
'procrustes': procrustes_error,
'knn_true': knn_error_true_emb,
'knn_ord_emb': knn_error_ord_emb,
'labels': labels,
'loss_history': loss_history,
'error_history': triplet_error_history,
'ordinal_embedding': x,
'time_taken': time_taken
}
joblib.dump(results, os.path.join(log_dir, experiment_name + '.pkl'))
def run_hyper_search(config):
"""
Important Hyperparameters for tSTE:
Learning Rate: [1, 0.1, 0.01]
"""
dataset_name = config['dataset_selected']
batch_size = config['batch_size']
epochs = config['nb_epochs']
input_dim = config['input_dimension']
n_samples = config['number_of_points']
number_of_test_triplets = config['n_test_triplets']
log_dir = config['log']['path']
triplet_multiplier_range = config['hyper_search']['triplets_multiplier']
learning_rate_range = config['hyper_search']['learning_rate']
optimizer = config['optimizer']
dimensions_range = config['hyper_search']['output_dimension']
separator = '_'
experiment_name = 'tste_hyper_search_' + \
'data_' + dataset_name + \
'_input_dim_' + str(input_dim) + \
'_n_pts_' + str(n_samples) + \
'_num_test_trips_' + str(number_of_test_triplets) + \
'_output_dim_' + separator.join([str(i) for i in dimensions_range]) + \
'_lr_' + separator.join([str(i) for i in learning_rate_range]) + \
'_optimizer_' + str(optimizer) + \
'_bs_' + str(batch_size) + \
'_triplet_number_' + separator.join([str(i) for i in triplet_multiplier_range])
if not os.path.exists(log_dir):
os.makedirs(log_dir)
logging_path = os.path.join(log_dir, experiment_name + '.log')
logger = logging_util.my_custom_logger(logger_name=logging_path,
level=logging.INFO)
logger.info('Name of Experiment: ' + experiment_name)
logger.info('Logging Path:' + logging_path)
logger.info('Dataset Name: ' + dataset_name)
logger.info('Epochs: ' + str(epochs))
best_params_train = {}
best_params_test = {}
all_results = {}
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
vec_data, labels = select_dataset(
dataset_name, n_samples=n_samples, input_dim=input_dim
) # input_dim is only argument for uniform. Ignored otherwise
n = vec_data.shape[0]
logn = int(np.log2(n))
for (emb_dim, triplet_multiplier) in product(dimensions_range,
triplet_multiplier_range):
all_results[(emb_dim, triplet_multiplier)] = {}
best_train_error = 1
best_test_error = 1
triplet_num = triplet_multiplier * logn * n * emb_dim
bs = min(batch_size, triplet_num)
train_triplets_dataset = TripletBatchesDataset(vec_data, labels,
triplet_num, bs, device)
logger.info('Testing on: ' + dataset_name +
'. Embedding dimension is ' + str(emb_dim))
logger.info(' ')
for learning_rate in learning_rate_range:
logger.info(10 * '-' + ' New parameters' + 10 * '-')
logger.info('Learning Rate: ' + str(learning_rate))
logger.info('Number of Points: ' + str(n))
logger.info('Input Dimension: ' + str(input_dim))
logger.info('Output Dimension: ' + str(emb_dim))
logger.info('Number of Test Triplets: ' +
str(number_of_test_triplets))
logger.info('Triplet Multiplier: ' + str(triplet_multiplier))
logger.info('Batch Size: ' + str(batch_size))
logger.info('Computing tSTE...')
x, loss_history, triplet_error_history, time_taken, time_history = tste.t_ste_adam(
triplets=train_triplets_dataset.trips_data_indices,
n=n,
emb_dim=emb_dim,
epochs=epochs,
batch_size=bs,
learning_rate=learning_rate,
device=device,
logger=logger)
# compute triplet error for train and test data
train_error = train_triplets_dataset.triplet_error(x)
logger.info('Triplet Error on Training Triplets: ' +
str(train_error))
test_triplets_dataset = TripletBatchesDataset(
vec_data, labels, number_of_test_triplets, 1000, device)
test_error = test_triplets_dataset.triplet_error(x)
#procrustes_error = procrustes_disparity(vec_data, x)
#knn_error_ord_emb, knn_error_true_emb = knn_classification_error(x, vec_data, labels)
logger.info('Epochs: ' + str(epochs))
logger.info('Time Taken: ' + str(time_taken) + ' seconds.')
logger.info('Train Error: ' + str(train_error))
logger.info('Test Error: ' + str(test_error))
#logger.info('Procrustes Disparity: ' + str(procrustes_error))
#logger.info('kNN Classification Error on ground-truth: ' + str(knn_error_true_emb))
#logger.info('kNN Classification Error on embedding: ' + str(knn_error_ord_emb))
results = {
'train_error': train_error,
'test_error': test_error,
'loss_history': loss_history,
'error_history': triplet_error_history,
'last_embedding': x
}
all_results[(emb_dim,
triplet_multiplier)].update({learning_rate: results})
if test_error < best_test_error:
best_params_test[(emb_dim, triplet_multiplier)] = {
'learning_rate': learning_rate,
'optimizer': optimizer,
'error': test_error
}
best_test_error = test_error
if train_error < best_train_error:
best_params_train[(emb_dim, triplet_multiplier)] = {
'learning_rate': learning_rate,
'optimizer': optimizer,
'error': train_error
}
best_train_error = train_error
result_name = 'tste_convergence_' + \
'data_' + dataset_name + \
'_input_dim_' + str(input_dim) + \
'_n_pts_' + str(n_samples) + \
'_output_dim_' + str(emb_dim) + \
'_bs_' + str(batch_size) + \
'_triplet_number_' + str(triplet_multiplier)
all_results['labels'] = labels
joblib.dump(all_results[(emb_dim, triplet_multiplier)],
os.path.join(log_dir, result_name + '.pkl'))
# print all results as well again
logger.info(10 * '-' + 'ALL RESULTS ' + 10 * '-')
for (emb_dim, triplet_multiplier) in product(dimensions_range,
triplet_multiplier_range):
results = all_results[(emb_dim, triplet_multiplier)]
logger.info('Results for emb dimension ' + str(emb_dim) +
' and triplet multiplier ' + str(triplet_multiplier))
for learning_rate in learning_rate_range:
logger.info('learning rate ' + str(learning_rate) +
' -- train error: ' +
str(results[learning_rate]['train_error']) +
' test error: ' +
str(results[learning_rate]['test_error']))
# print best parameter settings
for (emb_dim, triplet_multiplier) in product(dimensions_range,
triplet_multiplier_range):
logger.info('Best Parameters for emb dimension ' + str(emb_dim) +
' and triplet multiplier ' + str(triplet_multiplier))
best_on_train = best_params_train[(emb_dim, triplet_multiplier)]
best_on_test = best_params_test[(emb_dim, triplet_multiplier)]
logger.info('achieved ' + str(best_on_train['error']) +
' train error with learning rate: ' +
str(best_on_train['learning_rate']))
logger.info('achieved ' + str(best_on_test['error']) +
' test error with learning rate: ' +
str(best_on_test['learning_rate']))
def main(args):
config = load_config(args.config_path)
dataset_name = config['dataset_selected']
batch_size = config['batch_size']
phase1_learning_rate = config['optimizer_params']['phase1_learning_rate']
phase2_learning_rate = config['optimizer_params']['phase2_learning_rate']
num_landmarks = config['optimizer_params']['num_landmarks']
subset_size = config['optimizer_params']['subset_size']
target_loss = config['optimizer_params']['target_loss']
epochs = config['nb_epochs']
input_dim = config['input_dimension']
embedding_dimension = config['output_dimension']
n_samples = config['number_of_points']
number_of_test_triplets = config['n_test_triplets']
log_dir = config['log']['path']
hyper_search = config['hyper_search']['activation']
if hyper_search:
run_hyper_search(config=config)
else:
vec_data, labels = select_dataset(dataset_name=dataset_name,
input_dim=input_dim, n_samples=n_samples)
n_points = vec_data.shape[0] # do not remove
number_of_landmarks = min(int(num_landmarks * n_points), 100)
subset_size = subset_size * number_of_landmarks
experiment_name = 'lsoe_' + 'data_' + dataset_name \
+ '_input_dim_' + str(input_dim) \
+ '_emb_dimension_' + str(embedding_dimension) \
+ '_originaldimension_' + str(vec_data.shape[1]) \
+ '_n_' + str(n_samples) \
+ '_landmarks_' + str(number_of_landmarks) \
+ '_bs_ ' + str(batch_size) \
+ '_pplr_' + str(phase2_learning_rate) \
+ '_soe_lr_' + str(phase1_learning_rate) \
+ '_epochs_' + str(epochs)
if not os.path.exists(log_dir):
os.makedirs(log_dir)
logging_path = os.path.join(log_dir, experiment_name + '.log')
logger = logging_util.my_custom_logger(logger_name=logging_path, level=logging.INFO)
logger.info('Name of Experiments: ' + experiment_name)
logger.info('Dataset Name:' + dataset_name)
logger.info('Number of Points: ' + str(n_samples))
logger.info('Dataset Dimension:' + str(input_dim))
logger.info('Number of Landmarks:' + str(number_of_landmarks))
logger.info('Number of Subset Size:' + str(subset_size))
logger.info('First Phase Epochs: ' + str(epochs))
# set the gpu id
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
logger.info('Computing SOE - Phase 1...')
# first phase of the algorithm
landmarks, first_phase_indices, \
first_phase_subset_size, first_phase_reconstruction, \
first_phase_loss, first_phase_triplet_error, time_first_phase = first_phase_soe(
num_landmarks=number_of_landmarks,
subset_size=subset_size,
data=vec_data, dataset_size=n_points,
embedding_dim=embedding_dimension, epochs=epochs,
first_phase_lr=phase1_learning_rate,
device=device,
target_loss=target_loss,
batch_size=batch_size,
logger=logger)
logger.info('First Phase Loss: ' + str(first_phase_loss))
logger.info('First Phase Triplet Error: ' + str(first_phase_triplet_error))
logger.info('First Phase Number of Landmarks: ' + str(landmarks.shape))
logger.info('First Phase Indices Number: ' + str(len(first_phase_indices)))
logger.info('First Phase Reconstruction Size: ' + str(first_phase_reconstruction.shape))
embedded_indices = first_phase_indices
embedded_points = first_phase_reconstruction
non_embedded_indices = list(set(range(vec_data.shape[0])).difference(set(embedded_indices)))
my_oracle = Oracle(data=vec_data)
logger.info('Second Phase: ')
logger.info('Oracle Created...')
logger.info('Computing LLOE - Phase 2...')
# second phase for embedding point by point update
second_phase_embeddings_index, \
second_phase_embeddings, time_second_phase = second_phase(my_oracle=my_oracle,
non_embedded_indices=non_embedded_indices,
embedded_indices=embedded_indices,
first_phase_embedded_points=embedded_points,
dim=embedding_dimension,
lr=phase2_learning_rate, logger=logger)
# combine the first phase and second phase points and index
final_embedding = np.zeros((vec_data.shape[0], embedding_dimension))
# phase 1 points
final_embedding[embedded_indices] = embedded_points
# second phase points
final_embedding[second_phase_embeddings_index] = second_phase_embeddings
time_taken = time_first_phase + time_second_phase
logger.info('Size of Dataset: ' + str(vec_data.shape[0]))
logger.info('Size of First Phase Indices: ' + str(len(embedded_indices)))
logger.info('Size of Second Phase Indices: ' + str(len(second_phase_embeddings_index)))
# Evaluation
logger.info('Evaluation of the Complete Embedding Dataset: ')
random_trip_indices = gen_triplet_indices(n=vec_data.shape[0], num_trips=number_of_test_triplets)
test_triplet_data = gen_triplet_data(data=vec_data, random_triplet_indices=random_trip_indices, batch_size=1000)
test_error, embedding_error_list = triplet_error(final_embedding, test_triplet_data)
procrustes_error = procrustes_disparity(vec_data, final_embedding)
knn_error_ord_emb, knn_error_true_emb = knn_classification_error(final_embedding, vec_data, labels)
# sample points for tsne visualization
subsample = np.random.permutation(n_points)[0:500]
x = final_embedding[subsample, :]
sub_labels = labels[subsample]
x_embedded = TSNE(n_components=2, perplexity=15, learning_rate=10).fit_transform(x)
fig, ax = plt.subplots(1, 1)
ax.scatter(x_embedded[:, 0], x_embedded[:, 1], s=3, c=sub_labels)
fig.savefig(os.path.join(log_dir, experiment_name + '.png'))
logger.info('Name of Experiments: ' + experiment_name)
logger.info('Epochs: ' + str(epochs))
logger.info('Time Taken: ' + str(time_taken) + ' seconds.')
logger.info('Test Error: ' + str(test_error))
logger.info('Procrustes Disparity: ' + str(procrustes_error))
logger.info('kNN Classification Error on ground-truth: ' + str(knn_error_true_emb))
logger.info('kNN Classification Error on embedding: ' + str(knn_error_ord_emb))
results = {'test_error': test_error, 'procrustes': procrustes_error, 'knn_true': knn_error_true_emb,
'knn_ord_emb': knn_error_ord_emb, 'labels': labels,
'ordinal_embedding': final_embedding, 'time_taken': time_taken}
joblib.dump(results, os.path.join(log_dir, experiment_name + '.pkl'))
def main(args):
config = load_config(args.config_path)
algorithm = config['algorithm']
error_change_threshold = config['error_change_threshold']
input_dim_range = config['tradeoff_set']['input_dimension']
output_dimensions_range = config['tradeoff_set']['output_dimension']
nmb_points_range = config['tradeoff_set']['number_of_points']
batch_size_range = config['tradeoff_set']['batch_size']
learning_rate_range = config['tradeoff_set']['learning_rate']
triplet_multiplier_range = config['tradeoff_set']['triplets_multiplier']
if args.data_set == 'not_selected':
dataset_name = config['dataset_selected']
else:
dataset_name = args.data_set
try:
input_equals_output = config['tradeoff_set']['input_equals_output']
except:
input_equals_output = False
epochs = config['nb_epochs']
optimizer = config['optimizer']
n_test_triplets = config['n_test_triplets']
log_dir = config['log']['path']
separator = '_'
experiment_name = algorithm + \
'_data_' + dataset_name + \
'_input_dim_' + separator.join([str(i) for i in input_dim_range]) + \
'_output_dim_' + separator.join([str(i) for i in output_dimensions_range]) + \
'_n_pts_' + separator.join([str(i) for i in nmb_points_range]) + \
'_bs_' + separator.join([str(i) for i in batch_size_range]) + \
'_change_criterion_' + str(error_change_threshold) + \
'_lr_' + separator.join([str(i) for i in learning_rate_range]) + \
'_triplet_num_' + separator.join([str(i) for i in triplet_multiplier_range]) + \
'_ep_' + str(epochs)
# create a log directory if it does not exist
if not os.path.exists(log_dir):
os.makedirs(log_dir)
logging_path = os.path.join(log_dir, experiment_name + '.log')
logger = logging_util.my_custom_logger(logger_name=logging_path,
level=logging.INFO)
logger.info('Name of Experiments: ' + experiment_name)
logger.info('Dataset Name: ' + dataset_name)
logger.info('Error Change Threshold: ' + str(error_change_threshold))
logger.info('Epochs: ' + str(epochs))
tradeoff_results = defaultdict(dict)
order = [
'Input Dimension', 'Output Dimension', 'Number of Points',
'Batch Size', 'Learning Rate', 'Triplet Multiplier',
[
'Train Error', 'Test Error', 'Procrustes Error', 'Knn Orig Error',
'Knn Ordinal Error', 'Time', 'Embedding', 'Labels',
'Train Triplets'
]
]
experiment_range = OrderedDict({
'input_dim':
input_dim_range,
'output_dim':
output_dimensions_range,
'number_of_points':
nmb_points_range,
'batch_size':
batch_size_range,
'learning_rate':
learning_rate_range,
'triplet_multiplier':
triplet_multiplier_range
})
for input_dim_index, input_dim in enumerate(input_dim_range):
for dimensions_index, embedding_dimension in enumerate(
output_dimensions_range):
for subset_index, nmb_points in enumerate(nmb_points_range):
for batch_size_index, batch_size in enumerate(
batch_size_range):
for lr_index, learning_rate in enumerate(
learning_rate_range):
for trip_mindex, triplet_multiplier in enumerate(
triplet_multiplier_range):
if (not input_equals_output) or (
input_equals_output
and input_dim == embedding_dimension):
logger.info('Learning Rate: ' +
str(learning_rate))
logger.info('Number of Points: ' +
str(nmb_points))
logger.info('Input Dimension: ' +
str(input_dim))
logger.info('Output Dimension: ' +
str(embedding_dimension))
logger.info('Number of Test Triplets: ' +
str(n_test_triplets))
logger.info('Triplet Multiplier: ' +
str(triplet_multiplier))
logger.info('Batch Size: ' + str(batch_size))
embedding, train_triplets, labels, train_error, test_error, \
procrustes_error, knn_orig, knn_embed, time_taken, \
loss_history, triplet_error_history, time_history = run_method(config, dataset_name,
algorithm, nmb_points,
input_dim,
embedding_dimension,
learning_rate,
batch_size,
triplet_multiplier,
optimizer, epochs,
n_test_triplets, logger,
error_change_threshold)
tradeoff_results[input_dim_index,
dimensions_index,
subset_index,
batch_size_index, lr_index,
trip_mindex, 0] = train_error
tradeoff_results[input_dim_index,
dimensions_index,
subset_index,
batch_size_index, lr_index,
trip_mindex, 1] = test_error
tradeoff_results[input_dim_index,
dimensions_index,
subset_index,
batch_size_index, lr_index,
trip_mindex,
2] = procrustes_error
tradeoff_results[input_dim_index,
dimensions_index,
subset_index,
batch_size_index, lr_index,
trip_mindex, 3] = knn_orig
tradeoff_results[input_dim_index,
dimensions_index,
subset_index,
batch_size_index, lr_index,
trip_mindex, 4] = knn_embed
tradeoff_results[input_dim_index,
dimensions_index,
subset_index,
batch_size_index, lr_index,
trip_mindex, 5] = time_taken
tradeoff_results[input_dim_index,
dimensions_index,
subset_index,
batch_size_index, lr_index,
trip_mindex, 6] = embedding
tradeoff_results[input_dim_index,
dimensions_index,
subset_index,
batch_size_index, lr_index,
trip_mindex, 7] = labels
tradeoff_results[input_dim_index,
dimensions_index,
subset_index,
batch_size_index, lr_index,
trip_mindex, 8] = loss_history
tradeoff_results[input_dim_index,
dimensions_index,
subset_index,
batch_size_index, lr_index,
trip_mindex,
9] = triplet_error_history
tradeoff_results[input_dim_index,
dimensions_index,
subset_index,
batch_size_index, lr_index,
trip_mindex,
10] = time_history
# tradeoff_results[input_dim_index, dimensions_index, subset_index,
# batch_size_index, lr_index, trip_mindex, 8] = train_triplets
for input_dim_index, input_dim in enumerate(input_dim_range):
for dimensions_index, embedding_dimension in enumerate(
output_dimensions_range):
for subset_index, nmb_points in enumerate(nmb_points_range):
for batch_size_index, batch_size in enumerate(
batch_size_range):
for lr_index, learning_rate in enumerate(
learning_rate_range):
for trip_mindex, triplet_multiplier in enumerate(
triplet_multiplier_range):
if (not input_equals_output) or (
input_equals_output
and input_dim == embedding_dimension):
logger.info('Input Dimension ' +
str(input_dim) +
' Output Dimension ' +
str(embedding_dimension) +
' Number of Points ' +
str(nmb_points) + ' Batch Size ' +
str(batch_size) +
' Learning Rate ' +
str(learning_rate) +
' Triplet Multiplier ' +
str(triplet_multiplier))
logger.info(
' Train Error ' + str(tradeoff_results[
input_dim_index, dimensions_index,
subset_index, batch_size_index,
lr_index, trip_mindex, 0]) +
' Test Error ' + str(tradeoff_results[
input_dim_index, dimensions_index,
subset_index, batch_size_index,
lr_index, trip_mindex, 1]))
logger.info(' Procrustes Error ' +
str(tradeoff_results[
input_dim_index,
dimensions_index, subset_index,
batch_size_index, lr_index,
trip_mindex, 2]))
logger.info(
' kNN original Emb Loss ' +
str(tradeoff_results[
input_dim_index, dimensions_index,
subset_index, batch_size_index,
lr_index, trip_mindex, 3]) +
' kNN on Ordinal Emb Loss ' +
str(tradeoff_results[
input_dim_index, dimensions_index,
subset_index, batch_size_index,
lr_index, trip_mindex, 4]))
logger.info(' Time ' + str(tradeoff_results[
input_dim_index, dimensions_index,
subset_index, batch_size_index, lr_index,
trip_mindex, 5]))
logger.info('-' * 20)
data_dump = [order, experiment_range, tradeoff_results]
joblib.dump(data_dump, os.path.join(log_dir, experiment_name + '.pkl'))
def run_hyper_search(config):
dataset_name = config['dataset_selected']
batch_size = config['batch_size']
epochs = config['nb_epochs']
input_dim = config['input_dimension']
n_samples = config['number_of_points']
number_of_test_triplets = config['n_test_triplets']
log_dir = config['log']['path']
phase1_learning_rate_range = config['hyper_search']['phase1_learning_rate']
phase2_learning_rate_range = config['hyper_search']['phase2_learning_rate']
dimensions_range = config['hyper_search']['output_dimension']
separator = '_'
experiment_name = 'lloe_full_hyper_search_' + \
'data_' + dataset_name + \
'_input_dim_' + str(input_dim) + \
'_n_pts_' + str(n_samples) + \
'_num_test_trips_' + str(number_of_test_triplets) + \
'_output_dim_' + separator.join([str(i) for i in dimensions_range]) + \
'_phase1lr_' + separator.join([str(i) for i in phase1_learning_rate_range]) + \
'_phase2lr_' + separator.join([str(i) for i in phase2_learning_rate_range]) + \
'_bs_' + str(batch_size)
if not os.path.exists(log_dir):
os.makedirs(log_dir)
logging_path = os.path.join(log_dir, experiment_name + '.log')
logger = logging_util.my_custom_logger(logger_name=logging_path, level=logging.INFO)
logger.info('Name of Experiment: ' + experiment_name)
logger.info('Logging Path:' + logging_path)
logger.info('Dataset Name: ' + dataset_name)
logger.info('Epochs: ' + str(epochs))
best_params_test = {}
all_results = {}
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
vec_data, labels = select_dataset(dataset_name, n_samples=n_samples, input_dim=input_dim) # input_dim is only argument for uniform. Ignored otherwise
n_samples = vec_data.shape[0] # do not remove
number_of_landmarks = int(0.1 * n_samples)
subset_size = 10 * number_of_landmarks
for emb_dim in dimensions_range:
all_results[emb_dim] = {}
best_test_error = 1
logger.info('Testing on: ' + dataset_name + '. Embedding dimension is ' + str(emb_dim))
logger.info(' ')
for (phase1_learning_rate, phase2_learning_rate) \
in product(phase1_learning_rate_range, phase2_learning_rate_range):
logger.info(10*'-'+' New parameters' + 10*'-')
logger.info('phase1_Learning Rate: ' + str(phase1_learning_rate))
logger.info('phase2_Learning Rate: ' + str(phase2_learning_rate))
logger.info('Number of Points: ' + str(n_samples))
logger.info('Input Dimension: ' + str(input_dim))
logger.info('Output Dimension: ' + str(emb_dim))
logger.info('Number of Test Triplets: ' + str(number_of_test_triplets))
logger.info('Batch Size: ' + str(batch_size))
logger.info('Computing LOE_FULL...')
# first phase of the algorithm
landmarks, first_phase_indices, \
first_phase_subset_size, first_phase_reconstruction, \
first_phase_loss, first_phase_triplet_error, time_first_phase = first_phase_soe(num_landmarks=number_of_landmarks,
subset_size=subset_size,
data=vec_data, dataset_size=n_samples,
embedding_dim=emb_dim, epochs=epochs,
target_loss=0.1,
first_phase_lr=phase1_learning_rate,
device=device,
batch_size=batch_size, logger=logger)
logger.info('First Phase Loss: ' + str(first_phase_loss))
logger.info('First Phase Triplet Error: ' + str(first_phase_triplet_error))
logger.info('First Phase Number of Landmarks: ' + str(landmarks.shape))
logger.info('First Phase Indices Number: ' + str(len(first_phase_indices)))
logger.info('First Phase Reconstruction Size: ' + str(first_phase_reconstruction.shape))
embedded_indices = first_phase_indices
embedded_points = first_phase_reconstruction
non_embedded_indices = list(set(range(vec_data.shape[0])).difference(set(embedded_indices)))
non_embedded_points = vec_data[non_embedded_indices, :]
my_oracle = Oracle(data=vec_data)
logger.info('Second Phase: ')
logger.info('Oracle Created...')
logger.info('Computing LLOE - Phase 2...')
# second phase for embedding point by point update
seocnd_phase_embeddings_index, \
second_phase_embeddings, time_second_phase = second_phase(my_oracle=my_oracle,
non_embedded_indices=non_embedded_indices,
embedded_indices=embedded_indices,
first_phase_embedded_points=embedded_points,
dim=emb_dim,
lr=phase2_learning_rate, logger=logger)
# combine the first phase and second phase points and index
final_embedding = np.zeros((vec_data.shape[0], emb_dim))
# phase 1 points
final_embedding[first_phase_indices] = first_phase_reconstruction
# second phase points
final_embedding[seocnd_phase_embeddings_index] = second_phase_embeddings
logger.info('Size of First Phase Indices: ' + str(len(first_phase_indices)))
logger.info('Size of Second Phase Indices: ' + str(len(seocnd_phase_embeddings_index)))
logger.info('First Phase Triplet Error: ' + str(first_phase_triplet_error))
# Evaluation
logger.info('Evaluation of the Complete Embedding Dataset: ')
random_trip_indices = gen_triplet_indices(n=vec_data.shape[0], num_trips=number_of_test_triplets)
test_triplet_data = gen_triplet_data(data=vec_data, random_triplet_indices=random_trip_indices, batch_size=1000)
test_error, embedding_error_list = triplet_error(final_embedding, test_triplet_data)
time_taken = time_first_phase + time_second_phase
logger.info('Time Taken: ' + str(time_taken) + ' seconds.')
logger.info('Test Error: ' + str(test_error))
#logger.info('Procrustes Disparity: ' + str(procrustes_error))
#logger.info('kNN Classification Error on ground-truth: ' + str(knn_error_true_emb))
#logger.info('kNN Classification Error on embedding: ' + str(knn_error_ord_emb))
results = {'test_error': test_error, 'last_embedding': final_embedding}
all_results[emb_dim].update({(phase1_learning_rate, phase2_learning_rate): results})
if test_error < best_test_error:
best_params_test[emb_dim] = {'phase1_learning_rate': phase1_learning_rate, 'phase2_learning_rate': phase2_learning_rate,
'error': test_error}
best_test_error = test_error
result_name = 'lloe_full_hypersearch_' + \
'data_' + dataset_name + \
'_input_dim_' + str(input_dim) + \
'_n_pts_' + str(n_samples) + \
'_output_dim_' + str(emb_dim) + \
'_bs_' + str(batch_size)
all_results['labels'] = labels
joblib.dump(all_results[emb_dim], os.path.join(log_dir, result_name + '.pkl'))
# print all results as well again
logger.info(10 * '-' + 'ALL RESULTS ' + 10 * '-')
for emb_dim in dimensions_range:
results = all_results[emb_dim]
logger.info('Results for emb dimension ' + str(emb_dim))
for (phase1_learning_rate, phase2_learning_rate) \
in product(phase1_learning_rate_range, phase2_learning_rate_range):
logger.info('phase1_learning_rate ' + str(phase1_learning_rate) + ' phase2_learning_rate ' + str(phase2_learning_rate)
+ ' -- test error: ' + str(results[(phase1_learning_rate, phase2_learning_rate)]['test_error']))
# print best parameter settings
for emb_dim in dimensions_range:
logger.info('Best Parameters for emb dimension ' + str(emb_dim))
best_on_test = best_params_test[emb_dim]
logger.info('achieved ' + str(best_on_test['error']) + ' test error with phase1_learning_rate: ' + str(best_on_test['phase1_learning_rate'])
+ ' phase2_learning_rate: ' + str(best_on_test['phase2_learning_rate']))
