def transformer_test(transformer):
hits_params = {
loader_keys.DATA_PATH:
os.path.join(PROJECT_PATH, '../datasets/HiTS2013_300k_samples.pkl'),
loader_keys.N_SAMPLES_BY_CLASS:
10000,
loader_keys.TEST_PERCENTAGE:
0.2,
loader_keys.VAL_SET_INLIER_PERCENTAGE:
0.1,
loader_keys.USED_CHANNELS: [0, 1, 2, 3],
loader_keys.CROP_SIZE:
21,
general_keys.RANDOM_SEED:
42,
loader_keys.TRANSFORMATION_INLIER_CLASS_VALUE:
1
}
hits_outlier_dataset = HiTSOutlierLoader(hits_params)
(x_train, y_train), (x_val, y_val), (
x_test,
y_test) = hits_outlier_dataset.get_outlier_detection_datasets()
x_train_transform, y_train_transform = transformer.apply_all_transforms(
x=x_train)
x_val_transform, y_val_transform = transformer.apply_all_transforms(
x=x_val)
x_test_transform, y_test_transform = transformer.apply_all_transforms(
x=x_test)
return x_train_transform, y_train_transform, x_val_transform, \
y_val_transform, x_test_transform, y_test_transform
def hits4c_tr18():
hits_params = {
loader_keys.DATA_PATH:
os.path.join(PROJECT_PATH, '../datasets/HiTS2013_300k_samples.pkl'),
loader_keys.N_SAMPLES_BY_CLASS:
10000,
loader_keys.TEST_PERCENTAGE:
0.2,
loader_keys.VAL_SET_INLIER_PERCENTAGE:
0.1,
loader_keys.USED_CHANNELS: [0, 1, 2, 3],
loader_keys.CROP_SIZE:
21,
general_keys.RANDOM_SEED:
42,
loader_keys.TRANSFORMATION_INLIER_CLASS_VALUE:
1
}
data_loader = HiTSOutlierLoader(hits_params)
(x_train, y_train), (x_val, y_val), (
x_test, y_test) = data_loader.get_outlier_detection_datasets()
transformer = transformations_tf.KernelTransformer(flips=True,
gauss=False,
log=False)
mdl = EnsembleOVOTransformODSimpleModel(
data_loader,
transformer=transformer,
input_shape=x_train.shape[1:],
results_folder_name='transform_selection_1')
mdl.fit(x_train, x_val, train_batch_size=1024, verbose=0)
train_matrix_scores = mdl.predict_matrix_score(x_train,
transform_batch_size=1024)
val_matrix_scores = mdl.predict_matrix_score(x_val,
transform_batch_size=1024)
test_outlier_matrix_scores = mdl.predict_matrix_score(
x_test[y_test == 0], transform_batch_size=1024)
utils.save_pickle(
train_matrix_scores,
os.path.join(
mdl.main_model_path,
'train_matrix_scores_translations+flip(18)_train_step.pkl'))
utils.save_pickle(
val_matrix_scores,
os.path.join(mdl.main_model_path,
'val_matrix_scores_translations+flip(18)_train_step.pkl'))
utils.save_pickle(
test_outlier_matrix_scores,
os.path.join(
mdl.main_model_path,
'test_matrix_scores_translations+flip(18)_train_step.pkl'))
def save_normal_and_transformed_data(
transformer,
normal_data_name='tf2_normal.pkl',
transformed_data_name='tf2_old_transformed.pkl'):
save_dir = os.path.join(PROJECT_PATH, 'tests', 'aux_data')
utils.check_paths(save_dir)
hits_params = {
loader_keys.DATA_PATH:
os.path.join(PROJECT_PATH, '../datasets/HiTS2013_300k_samples.pkl'),
loader_keys.N_SAMPLES_BY_CLASS:
10000,
loader_keys.TEST_PERCENTAGE:
0.2,
loader_keys.VAL_SET_INLIER_PERCENTAGE:
0.1,
loader_keys.USED_CHANNELS: [0, 1, 2, 3],
loader_keys.CROP_SIZE:
21,
general_keys.RANDOM_SEED:
42,
loader_keys.TRANSFORMATION_INLIER_CLASS_VALUE:
1
}
hits_outlier_dataset = HiTSOutlierLoader(hits_params)
(x_train, y_train), (x_val, y_val), (
x_test,
y_test) = hits_outlier_dataset.get_outlier_detection_datasets()
x_train_transform, y_train_transform = transformer.apply_all_transforms(
x=x_train)
x_val_transform, y_val_transform = transformer.apply_all_transforms(
x=x_val)
x_test_transform, y_test_transform = transformer.apply_all_transforms(
x=x_test)
normal_data = ((x_train, y_train), (x_val, y_val), (x_test, y_test))
transformed_data = ((x_train_transform, y_train_transform),
(x_val_transform, y_val_transform), (x_test_transform,
y_test_transform))
utils.save_pickle(normal_data, os.path.join(save_dir, normal_data_name))
utils.save_pickle(transformed_data,
os.path.join(save_dir, transformed_data_name))
def test_all_tf2_dh_transtransformer_1c():
transformer = transformations_tf.TransTransformer()
hits_params = {
loader_keys.DATA_PATH:
os.path.join(PROJECT_PATH, '../datasets/HiTS2013_300k_samples.pkl'),
loader_keys.N_SAMPLES_BY_CLASS:
10000,
loader_keys.TEST_PERCENTAGE:
0.2,
loader_keys.VAL_SET_INLIER_PERCENTAGE:
0.1,
loader_keys.USED_CHANNELS: [2],
loader_keys.CROP_SIZE:
21,
general_keys.RANDOM_SEED:
42,
loader_keys.TRANSFORMATION_INLIER_CLASS_VALUE:
1
}
hits_outlier_dataset = HiTSOutlierLoader(hits_params)
scores_list = []
delta_times_list = []
for i in range(N_RUNS):
mdl = TransformODSimpleModel(data_loader=None,
transformer=transformer,
input_shape=(21, 21, 1),
drop_rate=0.0)
start_time = time.time()
scores = test_all_tf2(mdl,
transformer,
hits_outlier_dataset,
dataset_name='hits-1-c',
epochs=2)
end_time = time.time()
delta_times_list.append(end_time - start_time)
scores_list.append(scores)
del mdl
file_path = os.path.join(PROJECT_PATH, 'tests', 'aux_results',
'test_models_tf1-tf2.txt')
print_scores_times_to_file(
file_path,
'all_tf2_unchanged_dh_transtransformer_1c_DP0.0_fast_no_prints\n NRUNS: %i'
% N_RUNS, scores_list, delta_times_list)
os.path.join(PROJECT_PATH, '../datasets/HiTS2013_300k_samples.pkl'),
loader_keys.N_SAMPLES_BY_CLASS:
10000,
loader_keys.TEST_PERCENTAGE:
0.2,
loader_keys.VAL_SET_INLIER_PERCENTAGE:
0.1,
loader_keys.USED_CHANNELS: [2],
loader_keys.CROP_SIZE:
21,
general_keys.RANDOM_SEED:
42,
loader_keys.TRANSFORMATION_INLIER_CLASS_VALUE:
1
}
data_loader = HiTSOutlierLoader(hits_params)
(x_train, y_train), (x_val, y_val), (
x_test, y_test) = data_loader.get_outlier_detection_datasets()
# transformer = transformations_tf.KernelTransformer(
# flips=True, gauss=False, log=False)
transformer = transformations_tf.Transformer()
x_train_transformed, transformations_inds = transformer.apply_all_transforms(
x_train)
x_val_transformed, transformations_inds_val = transformer.apply_all_transforms(
x_val)
mdl = DeepHits(input_shape=x_train.shape[1:],
n_classes=transformer.n_transforms)
mdl.fit(x_train_transformed,
to_categorical(transformations_inds),
verbose=1,
epochs=100,
os.path.join(PROJECT_PATH, '../datasets/HiTS2013_300k_samples.pkl'),
loader_keys.N_SAMPLES_BY_CLASS:
10000,
loader_keys.TEST_PERCENTAGE:
0.2,
loader_keys.VAL_SET_INLIER_PERCENTAGE:
0.1,
loader_keys.USED_CHANNELS: [0, 1, 2, 3],
loader_keys.CROP_SIZE:
21,
general_keys.RANDOM_SEED:
42,
loader_keys.TRANSFORMATION_INLIER_CLASS_VALUE:
1
}
hits_outlier_dataset = HiTSOutlierLoader(hits_params)
transformer = Transformer()
trans_transformer = TransTransformer()
kernel_transformer = KernelTransformer()
kernel_plus_transformer = PlusKernelTransformer()
# data_loader, transformer, dataset_name, class_idx_to_run_experiments_on, n_runs
# TODO: delgate names to data_laoders
experiments_list = [
# (
# ztf_outlier_dataset, trans_transformer, 'ztf-real-bog-v1', 'real',
# N_RUNS),
# (
# ztf_outlier_dataset, kernel_transformer, 'ztf-real-bog-v1', 'real',
# N_RUNS),
# (
# ztf_outlier_dataset, transformer, 'ztf-real-bog-v1', 'real',
os.path.join(PROJECT_PATH, '../datasets/HiTS2013_300k_samples.pkl'),
loader_keys.N_SAMPLES_BY_CLASS:
10000,
loader_keys.TEST_PERCENTAGE:
0.2,
loader_keys.VAL_SET_INLIER_PERCENTAGE:
0.1,
loader_keys.USED_CHANNELS: [0, 1, 2, 3],
loader_keys.CROP_SIZE:
21,
general_keys.RANDOM_SEED:
42,
loader_keys.TRANSFORMATION_INLIER_CLASS_VALUE:
1
}
hits_outlier_dataset = HiTSOutlierLoader(hits_params)
(x_train, y_train), (x_val, y_val), (
x_test,
y_test) = hits_outlier_dataset.get_outlier_detection_datasets()
transformer = TransTransformer()
x_train_transform, y_train_transform = transformer.apply_all_transforms(
x=x_train)
x_val_transform, y_val_transform = transformer.apply_all_transforms(
x=x_val)
x_test_transform, y_test_transform = transformer.apply_all_transforms(
x=x_test)
time.sleep(10)
mdl = AlreadyTransformODModel(transformer=transformer,
input_shape=x_train.shape[1:])
def best_score_evaluation(result_folder_name, epochs, patience=0):
trainer_params = {
param_keys.RESULTS_FOLDER_NAME: result_folder_name,
'epochs': epochs,
'patience': patience,
}
# data loaders
hits_params = {
loader_keys.DATA_PATH:
os.path.join(PROJECT_PATH, '../datasets/HiTS2013_300k_samples.pkl'),
loader_keys.N_SAMPLES_BY_CLASS:
10000,
loader_keys.TEST_PERCENTAGE:
0.2,
loader_keys.VAL_SET_INLIER_PERCENTAGE:
0.1,
loader_keys.USED_CHANNELS: [0, 1, 2, 3], # [2],#
loader_keys.CROP_SIZE:
21,
general_keys.RANDOM_SEED:
42,
loader_keys.TRANSFORMATION_INLIER_CLASS_VALUE:
1
}
hits_loader = HiTSOutlierLoader(hits_params)
ztf_params = {
loader_keys.DATA_PATH:
os.path.join(PROJECT_PATH, '../datasets/ztf_v1_bogus_added.pkl'),
loader_keys.VAL_SET_INLIER_PERCENTAGE:
0.1,
loader_keys.USED_CHANNELS: [0, 1, 2],
loader_keys.CROP_SIZE:
21,
general_keys.RANDOM_SEED:
42,
loader_keys.TRANSFORMATION_INLIER_CLASS_VALUE:
1
}
ztf_loader = ZTFOutlierLoader(ztf_params)
# transformers
transformer_72 = transformations_tf.Transformer()
trans_transformer = transformations_tf.TransTransformer()
kernel_transformer = transformations_tf.KernelTransformer()
plus_kernel_transformer = transformations_tf.PlusKernelTransformer()
# trainers
hits_trainer = Trainer(hits_loader, trainer_params)
ztf_trainer = Trainer(ztf_loader, trainer_params)
model_constructors_list = (TransformODModel, )
transformers_list = (
plus_kernel_transformer,
kernel_transformer,
transformer_72,
trans_transformer,
)
trainers_list = (hits_trainer, ) # (ztf_trainer, hits_trainer, )
trainer_model_transformer_tuples = list(
itertools.product(trainers_list, model_constructors_list,
transformers_list))
for trainer, model_constructor, transformer in trainer_model_transformer_tuples:
trainer.train_model_n_times(model_constructor,
transformer,
trainer_params,
train_times=TRAIN_TIME)
hits_trainer.create_tables_of_results_folders()
os.path.join(PROJECT_PATH, '../datasets/HiTS2013_300k_samples.pkl'),
loader_keys.N_SAMPLES_BY_CLASS:
10000,
loader_keys.TEST_PERCENTAGE:
0.2,
loader_keys.VAL_SET_INLIER_PERCENTAGE:
0.1,
loader_keys.USED_CHANNELS: [0, 1, 2, 3], # [2],#
loader_keys.CROP_SIZE:
21,
general_keys.RANDOM_SEED:
42,
loader_keys.TRANSFORMATION_INLIER_CLASS_VALUE:
1
}
data_loader = HiTSOutlierLoader(hits_params)
transformer = TransTransformer()
params = {
param_keys.RESULTS_FOLDER_NAME: 'trainer_test',
'epochs': 1,
}
trainer = Trainer(params)
trainer.train_model_n_times(TransformODModel,
data_loader,
transformer,
params,
train_times=3)
trainer.train_model_n_times(TransformODSimpleModel,
data_loader,
transformer,
params,