X_test = X[n / 2 :]
Y_test = Y[n / 2 :]
print ("Corr(X)")
print (np.round(np.corrcoef(X.T), 2))
print ("Corr(Y)")
print (np.round(np.corrcoef(Y.T), 2))
###############################################################################
# Canonical (symetric) PLS
# Transform data
# ~~~~~~~~~~~~~~
plsca = PLSCanonical(n_components=2)
plsca.fit(X_train, Y_train)
X_train_r, Y_train_r = plsca.transform(X_train, Y_train)
X_test_r, Y_test_r = plsca.transform(X_test, Y_test)
# Scatter plot of scores
# ~~~~~~~~~~~~~~~~~~~~~~
# 1) On diagonal plot X vs Y scores on each components
pl.figure(figsize=(12, 8))
pl.subplot(221)
pl.plot(X_train_r[:, 0], Y_train_r[:, 0], "ob", label="train")
pl.plot(X_test_r[:, 0], Y_test_r[:, 0], "or", label="test")
pl.xlabel("x scores")
pl.ylabel("y scores")
pl.title("Comp. 1: X vs Y (test corr = %.2f)" % np.corrcoef(X_test_r[:, 0], Y_test_r[:, 0])[0, 1])
pl.xticks(())
pl.yticks(())
pl.legend(loc="best")
# remove features that have NaNs, but keep track of them for the future
idx = np.isnan(X_meg).any(axis=0) # mask of features with at least 1 NaN
X_meg = X_meg[:,~idx]
print 'MEG Features left: %d/%d'%(X_meg.shape[1],len(idx))
X_fmri = preprocessing.scale(X_fmri)
X_meg = preprocessing.scale(X_meg)
y = inatt
from sklearn.pls import PLSCanonical
ncomps = 10
plsca = PLSCanonical(n_components=ncomps)
plsca.fit(X_meg, X_fmri)
X_mc, X_fc = plsca.transform(X_meg, X_fmri)
res = []
print seed, band
for comp in range(ncomps):
r,p = stats.pearsonr(X_mc[:,comp], y)
res += [r,p]
r,p = stats.pearsonr(X_fc[:,comp], y)
res += [r,p]
all_results.append(['%s_%d-%d'%(seed,band[0],band[1])] + res)
header = []
for d in range(ncomps):
header+=['meg r%d'%d, 'meg_p%d'%d, 'fmri r%d'%d, 'fmri_p%d'%d]
header = ['data'] + header
all_results.insert(0, header)
X_test = X[n / 2:]
Y_test = Y[n / 2:]
print "Corr(X)"
print np.round(np.corrcoef(X.T), 2)
print "Corr(Y)"
print np.round(np.corrcoef(Y.T), 2)
###############################################################################
# Canonical (symetric) PLS
# Transform data
# ~~~~~~~~~~~~~~
plsca = PLSCanonical(n_components=2)
plsca.fit(X_train, Y_train)
X_train_r, Y_train_r = plsca.transform(X_train, Y_train)
X_test_r, Y_test_r = plsca.transform(X_test, Y_test)
# Scatter plot of scores
# ~~~~~~~~~~~~~~~~~~~~~~
# 1) On diagonal plot X vs Y scores on each components
pl.subplot(221)
pl.plot(X_train_r[:, 0], Y_train_r[:, 0], "ob", label="train")
pl.plot(X_test_r[:, 0], Y_test_r[:, 0], "or", label="test")
pl.xlabel("x scores")
pl.ylabel("y scores")
pl.title('Comp. 1: X vs Y (test corr = %.2f)' %
np.corrcoef(X_test_r[:, 0], Y_test_r[:, 0])[0, 1])
pl.legend()
pl.subplot(224)
def siPLS(X, y):
plsca = PLSCanonical(n_components=2)
plsca.fit(X, y)
X_r = plsca.transform(X)
print X_r
