def get_dummy_sample(preprocessing=["microvolt_scaling", "filtering"]):
train_sample, valid_sample, test_sample = get_epochs_data(
train_subjects=[0],
valid_subjects=[1],
test_subjects=[2],
recording=[1],
crop_wake_mins=0,
preprocessing=preprocessing)
# for i in range(len(train_sample)):
# train_sample[i] = (train_sample[i][0][:50], train_sample[i][1],
# train_sample[i][2])
# for i in range(len(test_sample)):
# test_sample[i] = (test_sample[i][0][:50],
# test_sample[i][1], test_sample[i][2])
test_choice = np.random.choice(range(len(test_sample)),
size=2,
replace=False)
valid_choice = np.random.choice(range(len(test_sample)),
size=2,
replace=False)
train_tinying_dict = {0: [350, 1029, 1291, 1650, 1571]}
test_tinying_dict = {0: valid_choice}
valid_tinying_dict = {0: test_choice}
train_sample.tinying_dataset(train_tinying_dict)
test_sample.tinying_dataset(test_tinying_dict)
valid_sample.tinying_dataset(valid_tinying_dict)
return train_sample, valid_sample, test_sample
# train_sample, valid_sample, test_sample = get_epochs_data(
# train_subjects=range(0, 10), valid_subjects=range(10, 15),
# test_subjects=range(15, 25),
# preprocessing=["scaling"])
# dl_dataset_args["transform_type"] = "raw (no transforms)" \
# "+ scaling"
# main_compute([shallow_args, sleepstager_args],
# [dl_dataset_args, dl_dataset_args],
# train_sample, valid_sample, test_sample,
# sample_size_list, saving_params)
train_sample, valid_sample, test_sample = get_epochs_data(
train_subjects=range(0, 10), valid_subjects=range(10, 15),
test_subjects=range(15, 25),
preprocessing=["microvolt_scaling", "filtering"])
dl_dataset_args_with_transforms["transform_list"] = [
["add_noise_to_signal"]]
for magnitude in [0, 0.2, 0.4, 0.6, 0.8, 1, 2, 3]:
transforms_args["magnitude"] = magnitude
dl_dataset_args_with_transforms["transform_type"] = "gaussian noise, "\
"scaling, filtering" \
"+ magnitude : " + str(magnitude)
main_compute([sleepstager_args], [dl_dataset_args_with_transforms],
transforms_args, train_sample, valid_sample, test_sample,
sample_size_list, saving_params)
# dl_dataset_args["transform_type"] = "raw (no transforms)" \
# sample_size_list, saving_params)
# train_sample, valid_sample, test_sample = get_epochs_data(
# train_subjects=range(0, 10), valid_subjects=range(10, 15),
# test_subjects=range(15, 25),
# preprocessing=["scaling"])
# dl_dataset_args["transform_type"] = "raw (no transforms)" \
# "+ scaling"
# main_compute([shallow_args, sleepstager_args],
# [dl_dataset_args, dl_dataset_args],
# train_sample, valid_sample, test_sample,
# sample_size_list, saving_params)
train_sample, valid_sample, test_sample = get_epochs_data(
crop_wake_mins=30)
dl_dataset_args_with_transforms["transform_list"] = [["identity"]]
dl_dataset_args_with_transforms["transform_type"] = "Baseline"
main_compute([sleepstager_args], [dl_dataset_args_with_transforms],
transforms_args, train_sample, valid_sample, test_sample,
sample_size_list, saving_params)
dl_dataset_args_with_transforms["transform_list"] = [["randaugment"]]
transforms_args["n_transf"] = 2
transforms_args["magnitude"] = 0.4
dl_dataset_args_with_transforms["transform_type"] = "Randaugment"
main_compute([sleepstager_args], [dl_dataset_args_with_transforms],
transforms_args, train_sample, valid_sample, test_sample,
sample_size_list, saving_params)