def get_single_subj_dataset(self, subject=None, final_evaluation=False):
if subject not in self.datasets['train'].keys():
self.datasets['train'][subject], self.datasets['valid'][subject], self.datasets['test'][subject] = \
get_train_val_test(global_vars.get('data_folder'), subject)
single_subj_dataset = OrderedDict(
(('train', self.datasets['train'][subject]),
('valid', self.datasets['valid'][subject]),
('test', self.datasets['test'][subject])))
if final_evaluation:
single_subj_dataset['train'] = concatenate_sets(
[single_subj_dataset['train'], single_subj_dataset['valid']])
return single_subj_dataset
def get_single_subj_dataset(self, subject=None, final_evaluation=False):
if subject not in self.datasets['train'].keys():
self.datasets['train'][subject], self.datasets['valid'][subject], self.datasets['test'][subject] = \
get_train_val_test(global_vars.get('data_folder'), subject)
single_subj_dataset = OrderedDict((('train', self.datasets['train'][subject]),
('valid', self.datasets['valid'][subject]),
('test', self.datasets['test'][subject])))
if final_evaluation and global_vars.get('ensemble_iterations'):
single_subj_dataset['train'] = concatenate_sets(
[single_subj_dataset['train'], single_subj_dataset['valid']])
if global_vars.get('time_frequency'):
EEG_to_TF_mne(single_subj_dataset)
set_global_vars_by_dataset(single_subj_dataset['train'])
return single_subj_dataset
def target_exp(exp_name, csv_file, subjects, model_from_file=None, write_header=True):
stop_criterion, iterator, loss_function, monitors = get_settings()
if write_header:
with open(csv_file, 'a', newline='') as csvfile:
writer = csv.DictWriter(csvfile, fieldnames=FIELDNAMES)
writer.writeheader()
for subject_id in subjects:
train_set = {}
val_set = {}
test_set = {}
if model_from_file is not None and global_vars.get('per_subject_exclusive') and \
not_exclusively_in(subject_id, model_from_file):
continue
train_set[subject_id], val_set[subject_id], test_set[subject_id] =\
get_train_val_test(data_folder, subject_id)
eegnas_from_file = EEGNAS_from_file(iterator=iterator, exp_folder=f"results/{exp_name}", exp_name = exp_name,
train_set=train_set, val_set=val_set, test_set=test_set,
stop_criterion=stop_criterion, monitors=monitors, loss_function=loss_function,
subject_id=subject_id, fieldnames=FIELDNAMES,
csv_file=csv_file, model_from_file=model_from_file)
eegnas_from_file.run_target_model()
low_cut_hz = 0
fs = 250
valid_set_fraction = 0.2
dataset = 'BCI_IV_2a'
data_folder = '../data/'
global_vars.set('dataset', dataset)
set_params_by_dataset()
global_vars.set('cuda', True)
model_select = 'deep4'
model_dir = '143_x_evolution_layers_cross_subject'
model_name = 'best_model_9_8_6_7_2_1_3_4_5.th'
train_set = {}
val_set = {}
test_set = {}
train_set[subject_id], val_set[subject_id], test_set[subject_id] =\
get_train_val_test(data_folder, subject_id)
# Set if you want to use GPU
# You can also use torch.cuda.is_available() to determine if cuda is available on your machine.
cuda = True
set_random_seeds(seed=20170629, cuda=cuda)
# This will determine how many crops are processed in parallel
input_time_length = 450
# final_conv_length determines the size of the receptive field of the ConvNet
models = {
'evolution': torch.load(f'../models/{model_dir}/{model_name}'),
'deep4': target_model('deep')
}
model = models[model_select]
input_time_length = global_vars.get('input_time_len')
set_params_by_dataset()
model_selection = 'deep4'
cnn_layer = {'evolution': 10, 'deep4': 25}
filter_pos = {'evolution': 0, 'deep4': 0}
model_dir = '91_x_BCI_IV_2b'
model_name = 'best_model_5_1_8_7_9_2_3_4_6.th'
model = {
'evolution': torch.load(f'../../models/{model_dir}/{model_name}'),
'deep4': target_model('deep')
}
subject_id = 1
train_set = {}
val_set = {}
test_set = {}
train_set[subject_id], val_set[subject_id], test_set[subject_id] = \
get_train_val_test(global_vars.get('data_folder'), subject_id)
stop_criterion, iterator, loss_function, monitors = get_normal_settings()
naiveNAS = NaiveNAS(iterator=iterator,
exp_folder=None,
exp_name=None,
train_set=train_set,
val_set=val_set,
test_set=test_set,
stop_criterion=stop_criterion,
monitors=monitors,
loss_function=loss_function,
config=global_vars.config,
subject_id=subject_id,
fieldnames=None,
strategy='cross_subject',
evolution_file=None,