I think you are right. Updating k-means to be more effecient has a number
of flow-on effects like, as you say, improving EM in turn.
On 10 April 2014 02:58, Maheshakya Wijewardena <[email protected]>wrote:
> Hi,
>
> Currently in scikit-learn, Expectation maximization algorithm is used in
> K-means clustering model to determine optimal cluster centers and labels.
> In my opinion, the best place to apply LSH based ANN methods(proposed as a
> GSOC project) is at the E step of the EM algorithm. The assignments of each
> data point are determined at that step for the current setting of cluster
> centers.
> ANN search can be applied to find nearest cluster centers of each data
> point. In `sklearn.cluster.k_means_.py`, from `_labels_inertia` function,
> the assignments are calculated using `_assign_labels_array` and
> `_assign_labels_csr` functions. These functions choose the center with
> minimum euclidean distance. Instead of that, from an ANN search, nearest
> neighbors can be approximated.
>
> This is my current plan for this. Your feedback is welcome.
>
> Best regards,
> Maheshakya
> --
> Undergraduate,
> Department of Computer Science and Engineering,
> Faculty of Engineering.
> University of Moratuwa,
> Sri Lanka
>
>
> ------------------------------------------------------------------------------
> Put Bad Developers to Shame
> Dominate Development with Jenkins Continuous Integration
> Continuously Automate Build, Test & Deployment
> Start a new project now. Try Jenkins in the cloud.
> http://p.sf.net/sfu/13600_Cloudbees
> _______________________________________________
> Scikit-learn-general mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>
>
------------------------------------------------------------------------------
Put Bad Developers to Shame
Dominate Development with Jenkins Continuous Integration
Continuously Automate Build, Test & Deployment
Start a new project now. Try Jenkins in the cloud.
http://p.sf.net/sfu/13600_Cloudbees
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
