def regression_linear_ridge_modular(fm_train=traindat, fm_test=testdat, label_train=label_traindat, tau=1e-6):
from modshogun import RegressionLabels, RealFeatures
from modshogun import LinearRidgeRegression
rr = LinearRidgeRegression(tau, RealFeatures(traindat), RegressionLabels(label_train))
rr.train()
out = rr.apply(RealFeatures(fm_test)).get_labels()
return out, rr
def regression_linear_ridge_modular(fm_train=traindat,
fm_test=testdat,
label_train=label_traindat,
tau=1e-6):
from modshogun import RegressionLabels, RealFeatures
from modshogun import LinearRidgeRegression
rr = LinearRidgeRegression(tau, RealFeatures(traindat),
RegressionLabels(label_train))
rr.train()
out = rr.apply(RealFeatures(fm_test)).get_labels()
return out, rr
lsr.train(RealFeatures(X.T))
# gather LASSO path
path = np.zeros((p, LeastAngleRegression.get_path_size()))
for i in xrange(path.shape[1]):
path[:, i] = LeastAngleRegression.get_w(i)
evaluator = MeanSquaredError()
# apply on training data
mse_train = np.zeros(LeastAngleRegression.get_path_size())
for i in xrange(mse_train.shape[0]):
LeastAngleRegression.switch_w(i)
ypred = LeastAngleRegression.apply(RealFeatures(X.T))
mse_train[i] = evaluator.evaluate(ypred, RegressionLabels(y))
ypred = lsr.apply(RealFeatures(X.T))
mse_train_lsr = evaluator.evaluate(ypred, RegressionLabels(y))
# apply on test data
mse_test = np.zeros(LeastAngleRegression.get_path_size())
for i in xrange(mse_test.shape[0]):
LeastAngleRegression.switch_w(i)
ypred = LeastAngleRegression.apply(RealFeatures(Xtest.T))
mse_test[i] = evaluator.evaluate(ypred, RegressionLabels(y))
ypred = lsr.apply(RealFeatures(Xtest.T))
mse_test_lsr = evaluator.evaluate(ypred, RegressionLabels(y))
fig = plt.figure()
ax_path = fig.add_subplot(1, 2, 1)
plt.plot(xrange(path.shape[1]), path.T, ".-")
plt.legend(["%d" % (x + 1) for x in xrange(path.shape[0])])
lsr.train(RealFeatures(X.T))
# gather LASSO path
path = np.zeros((p, LeastAngleRegression.get_path_size()))
for i in xrange(path.shape[1]):
path[:, i] = LeastAngleRegression.get_w(i)
evaluator = MeanSquaredError()
# apply on training data
mse_train = np.zeros(LeastAngleRegression.get_path_size())
for i in xrange(mse_train.shape[0]):
LeastAngleRegression.switch_w(i)
ypred = LeastAngleRegression.apply(RealFeatures(X.T))
mse_train[i] = evaluator.evaluate(ypred, RegressionLabels(y))
ypred = lsr.apply(RealFeatures(X.T))
mse_train_lsr = evaluator.evaluate(ypred, RegressionLabels(y))
# apply on test data
mse_test = np.zeros(LeastAngleRegression.get_path_size())
for i in xrange(mse_test.shape[0]):
LeastAngleRegression.switch_w(i)
ypred = LeastAngleRegression.apply(RealFeatures(Xtest.T))
mse_test[i] = evaluator.evaluate(ypred, RegressionLabels(y))
ypred = lsr.apply(RealFeatures(Xtest.T))
mse_test_lsr = evaluator.evaluate(ypred, RegressionLabels(y))
fig = plt.figure()
ax_path = fig.add_subplot(1, 2, 1)
plt.plot(xrange(path.shape[1]), path.T, '.-')
plt.legend(['%d' % (x + 1) for x in xrange(path.shape[0])])
