"How can I control the feature demonsion using Kernel PCA?"

zhangcezhangce Member Posts: 2 Contributor I
edited May 2019 in Help

I'm using kernel pca in rapidminer, however I found no place to tune the demonsion of transformed space like PCA... i.e., after carried out Kernel PCA, I always get a space with # of demonsion equals to the # of sample datas...how can I tune it just PCA? thanks.

    <operator name="KernelPCA" class="KernelPCA">
        <parameter key="kernel_type" value="polynomial"/>
    </operator>
    <operator name="ModelApplier (2)" class="ModelApplier">
    </operator>

Answers

  • landland RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 2,531 Unicorn
    Hi,
    unfortunatly you have to perform a conventional feature selection after the kernelPCA, since it always returns the number of features it was possible to calculate. Unlike the PCA this new features are NOT ordered by their variance, prohibiting the usual heuristic of removing the least variant directions in data space.

    Greetings,
      Sebastian
  • zhangcezhangce Member Posts: 2 Contributor I

    Thank you very much~ :)~
Sign In or Register to comment.