Smola's method should be complete.

This commit is contained in:
Dongryeol Lee
2010-08-16 19:49:48 +00:00
parent 423dc84cb8
commit 5bb841a6cb
@@ -46,7 +46,11 @@ double SparseGreedyGprModel::PredictVariance(
Vector kernel_values;
ComputeKernelValues_(
covariance, point_indices_in_dictionary, point, &kernel_values);
Vector left_hand_side;
SolveSystem_(dictionary_for_error_, kernel_values, &left_hand_side);
double dot_product = la::Dot(left_hand_side, dot_product);
double self_correlation = covariance.NormSq(point);
return self_correlation - dot_product;
}
void SparseGreedyGprModel::FinalizeModel() {