def _compute_w_row(self, Xw, w, W, idx):
g, gd = self._exp(Xw.multiply(DenseMatrix(len(w), 1, w)))
w_new = column_means(elementwise_product(Xw, g, self.spark))
del g
w_new = w_new - gd * w
w_new = gs_decorrelate(w_new, W, idx)
w_new /= scipy.sqrt((w_new**2).sum())
return w_new
def _preprocess_data(self, data):
X = self._feature_matrix(data)
self.__means = column_means(X)
return RowMatrix(center(X, means=self.__means))
def _exp(X):
g = X.rows.map(lambda x: x * scipy.exp(-(scipy.power(x, 2.0)) / 2.0))
g_ = X.rows.map(lambda x: (1 - scipy.power(x, 2.0)) * scipy.exp(-(
scipy.power(x, 2.0)) / 2.0))
gm = column_means(g_).mean()
return RowMatrix(g), gm