def apply(self):
loss = [
"squared_loss", "huber", "epsilon_insensitive",
"squared_epsilon_insensitive"
][self.loss_function]
penalty = ["l1", "l2", "elasticnet"][self.penalty_type]
learning_rate = ["invscaling", "constant"][self.learning_rate]
common_args = dict(
loss=loss,
alpha=self.alpha,
epsilon=self.epsilon,
eta0=self.eta0,
l1_ratio=self.l1_ratio,
power_t=self.power_t,
penalty=penalty,
learning_rate=learning_rate,
n_iter=self.n_iter,
)
learner = linear.SGDRegressionLearner(preprocessors=self.preprocessors,
**common_args)
learner.name = self.learner_name
predictor = None
if self.data is not None:
predictor = learner(self.data)
predictor.name = self.learner_name
self.send("Learner", learner)
self.send("Predictor", predictor)
def test_SGDRegression(self):
nrows = 500
ncols = 5
x = np.sort(10 * np.random.rand(nrows, ncols))
y = np.sum(np.sin(x), axis=1).reshape(nrows, 1)
x1, x2 = np.split(x, 2)
y1, y2 = np.split(y, 2)
t = data.Table(x1, y1)
learn = lr.SGDRegressionLearner()
clf = learn(t)
z = clf(x2)
self.assertTrue((abs(z.reshape(-1, 1) - y2) < 4.0).all())