Inconsistencies for the kernel documentation #2901

Closed
arjoly opened this Issue Feb 26, 2014 · 3 comments

Comments

Projects
None yet
6 participants
Owner

arjoly commented Feb 26, 2014

Here some inconsistencies that I have encountered when I wanted to use kernel:

  • The degree In KernelPCA, svm-kernel and SVC, they don't agree with the presence of the parameter degree for the sigmoid kernel.parameter of KernelPCA is said to be used for the rbf kernel. However in the svm kernel documentation, there isn't any mention of the degree parameter.
  • In KernelPCA, svm-kernel and SVC, they don't agree with the presence of the parameter degree for the sigmoid kernel.
  • In the kernel pca narrative doc and in KernelPCA, there isn't any link to the kernel definition
  • When you browse the narrative doc in Pairwise metrics, Affinities and Kernels, there is no pointer to the kernel approximation, definitions of all kernels and estimators/transformers that can use kernels.

The information about kernels is spread in at least four page. It would be nice that Pairwise metrics, Affinities and Kernels contains most information with links to the relevant pages (and backlink from these page).

arjoly added this to the 0.15 milestone Feb 26, 2014

Contributor

matrixorz commented Mar 20, 2014

in KernelPCA, the sentence "Degree for poly, rbf and sigmoid kernels. Ignored by other kernels." is not correct, degree is just for poly,i have correct it ,that will correct the first and second problem.

@matrixorz matrixorz added a commit to matrixorz/scikit-learn that referenced this issue Mar 20, 2014

@matrixorz matrixorz fix issue #2901 08a7f67

@ogrisel ogrisel added a commit that referenced this issue Mar 20, 2014

@ogrisel ogrisel Merge pull request #2981 from matrixorz/master
fix issue #2901
d413229

ogrisel removed this from the 0.15 milestone Jun 4, 2014

amueller added this to the 0.15.1 milestone Jul 18, 2014

Contributor

zeppe commented Nov 30, 2014

In sklearn/metrics/pairwise.py the function additive_chi2_kernel doesn't define a valid kernel as it is not positive semi-definite. The correct version of the kernel should be 1+<current_version>.

Contributor

zeppe commented Nov 30, 2014

gave it a shot in #3916

larsmans closed this Dec 1, 2014

@jmetzen jmetzen added a commit to jmetzen/scikit-learn that referenced this issue Dec 13, 2014

@zeppe @jmetzen zeppe + jmetzen DOC fix kernels documentation inconsistencies b20895f
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment