Added some comments

--HG--
extra : convert_revision : svn%3Afdd8eb12-d10e-0410-9acb-85c331704f74/trunk%403635
This commit is contained in:
Davis King 2010-05-16 19:11:48 +00:00
parent 9b51f8dc00
commit 0d775acbe6
1 changed files with 4 additions and 1 deletions

View File

@ -193,6 +193,9 @@ void test_manifold_regularization (
// Now create the manifold regularizer. The result is a transformation matrix that // Now create the manifold regularizer. The result is a transformation matrix that
// embodies the manifold assumption discussed above. // embodies the manifold assumption discussed above.
linear_manifold_regularizer<sample_type> lmr; linear_manifold_regularizer<sample_type> lmr;
// use_gaussian_weights is a function object that tells lmr how to weight each edge. In this
// case we let the weight decay as edges get longer. So shorter edges are more important than
// longer edges.
lmr.build(samples, edges, use_gaussian_weights(0.1)); lmr.build(samples, edges, use_gaussian_weights(0.1));
const matrix<double> T = lmr.get_transformation_matrix(intrinsic_regularization_strength); const matrix<double> T = lmr.get_transformation_matrix(intrinsic_regularization_strength);
@ -203,7 +206,7 @@ void test_manifold_regularization (
// For convenience, generate a projection_function and merge the transformation // For convenience, generate a projection_function and merge the transformation
// matrix T into it. So proj(x) == T*ekm.project(x). // matrix T into it. That is, we will have: proj(x) == T*ekm.project(x).
projection_function<kernel_type> proj = ekm.get_projection_function(); projection_function<kernel_type> proj = ekm.get_projection_function();
proj.weights = T*proj.weights; proj.weights = T*proj.weights;