|
KVQ kpca object - Kernel Vector Quantization
|
A=KVQ(H) returns a kvq object initialized with hyperparameters H.
Hyperparameters, and their defaults
dist=1; -- allowed point to point distortion.
child=linear -- child stores the kernel. The kernel induced the used distance measure.
a.cutoff = 1./3. -- cutoff value for importance values
a.return_indices = 0 -- return indices of points instead of samples.
Model
a.keep -- kept data points -- store data points kept for model
a.alpha -- importance factors
Methods:
train, test
Example:
d=gen(toy('n=2'));
[r,a]=train(kvq,d)
plot(a,d)
|
Reference : A kernel approach vor vector quantization with guaranteed distortion bounds |
|
Author : M. Tipping, B. Schölkopf |