for fold in range(n_folds):
# weights explained: gat.estimator_[time_point][fold].steps[-1][-1].coef_
weights.append(
np.vstack([
gat.estimators_[idx][fold].steps[-1][-1].coef_
for idx in range(len(epochs.times))
]))
np.save(fname_weights, np.array(weights))
cov = mne.compute_covariance(epochs)
cov.save(fname_cov)
# define a layout
layout = mne.find_layout(epochs.info)
# additional properties
group_dict['layout'] = layout
group_dict['times'] = epochs.times
group_dict['sfreq'] = epochs.info['sfreq']
else:
group_dict = pickle.load(open(fname_group))
subjects = group_dict['subjects']
####################
# Group Statistics #
####################
group_dict.update(
group_stats(subjects, path, exp, filt, analysis, c_name,
reg_type=reg_type))
pickle.dump(group_dict, open(fname_group, 'w'))
train_times=train_times,
clf=clf,
cv=n_folds)
gat.fit(epochs, y=y)
gat.score(epochs, y=y)
np.save(fname_gat, gat.scores_)
# store weights
weights = list()
for fold in range(n_folds):
# weights explained: gat.estimator_[time_point][fold].steps[-1][-1].coef_
weights.append(
np.vstack([
gat.estimators_[idx][fold].steps[-1][-1].coef_
for idx in range(len(epochs.times))
]))
np.save(fname_weights, np.array(weights))
cov = mne.compute_covariance(epochs)
cov.save(fname_cov)
else:
group_dict = pickle.load(open(fname_group))
subjects = group_dict['subjects']
####################
# Group Statistics #
####################
group_dict = group_stats(subjects, path, exp, filt, analysis, c_names)
pickle.dump(group_dict, open(fname_group, 'w'))