Java Applet for Laplacian SVMs
Click
on "Load toy1". The large colored
points are labeled examples and the small white points are
unlabeled examples. Press "Run". You will see the decision
surface obtained by an SVM using just the labeled examples. Now change
the options in the text area below to "-t 2 -c 100 -l 1000" and press "Run".
This is the decision boundary obtained using Laplacian SVM. You
can upload your 2d dataset or interactively construct a dataset
by using the "class:1" button.
The toy examples demonstrate the "cluster" and the "manifold"
assumptions for semi-supervised learning. This applet has been adapted from the Libsvm package.
Options:
-t kernel_type : set type of kernel function (default 2)
0 -- linear: u'*v
1 -- polynomial: (gamma*u'*v + coef0)^degree
2 -- radial basis: exp(-gamma*|u-v|^2)
3 -- sigmoid: tanh(gamma*u'*v + coef0)
-d degree : set degree in polynomial kernel (default 1)
-g gamma : set gamma in radial basis kernel (default 1/k)
-r coef0 : set coef0 in polynomial/sigmoid kernel function
-c cost : set the parameter C of SVM (1/2*gamma_A)
For Semi-supervised learning, include the following parameters,
-l deformation parameter : ratio of intrinsic and extrinsic regularization (default 0: standard supervised SVM)
-k number of nearest neighbors used to contruct the Graph laplacian (default : 6)