else:
for j, gv.trial in enumerate(gv.trials):
X_S1 = X_epochs[j,0,:,:,i]
X_S2 = X_epochs[j,1,:,:,i]
X_S1_S2 = np.vstack([X_S1, X_S2])
coefs_boot.append(fct.bootstrap_clf_epoch(X_S1_S2, y_trials, clf, shuffle=0))
coefs_boot = np.array(coefs_boot)
print('coefs', coefs_boot.shape)
break
alpha_samples = []
cos_alp_samples = []
for sample in range(coefs_boot.shape[0]):
alpha, cos_alp = fct.get_cos(coefs_boot[sample])
alpha_samples.append(alpha)
cos_alp_samples.append(cos_alp)
alpha_samples = np.asarray(alpha_samples)
cos_alp_samples = np.asarray(cos_alp_samples)
print('samples', cos_alp_samples.shape)
mean_cos_samples = np.mean(cos_alp_samples,axis=0)
q1 = mean_cos_samples - np.percentile(cos_alp_samples, 25, axis=0)
q3 = np.percentile(cos_alp_samples, 75, axis=0) - mean_cos_samples
y_err = np.asarray([q1[1:],q3[1:]])
print('trial', gv.trial, 'cos_alp', mean_cos_samples, 'q1', q1, 'q3', q3)
print('X', X_trials.shape)
dC = []
for i in range(X_trials.shape[2]):
X = StandardScaler().fit_transform(X_trials[:,:,i])
kmeans = KMeans(n_clusters=2, n_init=100).fit(X)
cluster_center = kmeans.cluster_centers_
# print(cluster_center.shape)
dC.append(cluster_center[0]-cluster_center[1])
dC = np.asarray(dC)
# print(dC.shape)
alpha, cos_alp = fct.get_cos(dC)
print('trial', gv.trial, 'cos_alp', cos_alp)
# idx = np.where(X_trials<0)
# X = np.delete(X_trials, idx, axis=1)
# print('# trials, neurons, time')
# print('X', X.shape,'y', y_trials.shape)
# # for epoch in range(0, len(gv.epochs)):
# # dum = X[:,:,epoch]
# # X.append(StandardScaler().fit_transform(dum))
# # X = np.asarray(X)
# # X = np.rollaxis(X,2,1).transpose()
X_trials, y_trials = data.get_X_y_epochs(X_S1_trials, X_S2_trials)
print('X', X_trials.shape,'y', y_trials.shape)
if IF_GRID :
coefs, clf = fct.grid_search_cv_clf(X_trials, y_trials, shuffle=0, cv=10)
else :
X_detrend = []
for n_trial in range(0,X_trials.shape[0]):
fit_values = fct.detrend_data(X_trials[n_trial], poly_fit=1, degree=10)
X_detrend.append(fit_values)
X_detrend = np.asarray(X_detrend)
coefs = fct.coefs_clf(X_trials-X_detrend, y_trials, clf=clf)
print('coefs', coefs.shape)
alpha, cos_alp = fct.get_cos(coefs)
print('trial', gv.trial, 'cos_alp', cos_alp)
alpha_trials.append(alpha)
cos_alp_trials.append(cos_alp)
if IF_SHUFFLE:
mat_cos = []
for i in range(1000):
coefs_shuffle = fct.coefs_clf(X_trials, y_trials, clf=clf, shuffle=1)
alpha_shuffle, cos_alp_shuffle = fct.get_cos(coefs_shuffle)
mat_cos.append(cos_alp_shuffle)
mat_cos = np.asarray(mat_cos)