Kernel Density Estimation (kde)

[iglesias]

This is an entrance task for the fundamental machine learning algorithms GSoC project http://www.shogun-toolbox.org/page/Events/gsoc2014_ideas#fundamental.

• Implement a KDE class inside the distributions directory that performs kernel density estimation. This algorithm depends on the choice of a kernel and a nearest neighbours search algorithm.

Scikit-learn has a nice implementation of this algorithm, so feel free to draw inspiration from it: github.com/scikit-learn/scikit-learn/blob/master/sklearn/neighbors/kde.py

This task is fun because the algorithm is not particularly hard to implement and it is possible to make good-looking examples (ask @iglesias if you need an idea for this once the main algorithm and unit test are implemented).

[iglesias]

Feel free to write your name below if you are working on this task! This way we avoid wasting manpower on the same algorithm.

[Saurabh7]

for reference, working on this one. :)

[SanchitAggarwal]

I am attempting this :)

[Saurabh7]

@iglesias i figured i needed help on this: the scikit one computes log of density estimation ( not by default though) and uses log of the kernel for computation is that what would be expected here?

[karlnapf]

densities should always be evaluated in log-domain to avoid numerical problems

[cassiogreco]

I will attempt this as well

[arman-z]

Maybe I'm too late. But I'm also try to impelement it.

[vigsterkr]

@iglesias @mazumdarparijat this is done, or?

[iglesias]

2383