[
https://issues.apache.org/jira/browse/FLINK-2131?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14608434#comment-14608434
]
ASF GitHub Bot commented on FLINK-2131:
---------------------------------------
Github user sachingoel0101 commented on the pull request:
https://github.com/apache/flink/pull/757#issuecomment-117220314
Hey @thvasilo , I'm going to break up this PR further. The motivation is
that, the Sampling code should be available as a general feature. Given a
probability distribution over data, user should be able to sample as many
points as they want.
The Sampler will take the DataSet as input, number of samples required and
a function which determines the relative probability of a particular element
being picked, apart from specifying whether the elements should be sampled with
replacement or without replacement.
Let me know your thoughts. I'll work out a version in the meantime. If this
is desirable, I will file a JIRA ticket and open a separate PR.
> Add Initialization schemes for K-means clustering
> -------------------------------------------------
>
> Key: FLINK-2131
> URL: https://issues.apache.org/jira/browse/FLINK-2131
> Project: Flink
> Issue Type: Task
> Components: Machine Learning Library
> Reporter: Sachin Goel
> Assignee: Sachin Goel
>
> The Lloyd's [KMeans] algorithm takes initial centroids as its input. However,
> in case the user doesn't provide the initial centers, they may ask for a
> particular initialization scheme to be followed. The most commonly used are
> these:
> 1. Random initialization: Self-explanatory
> 2. kmeans++ initialization: http://ilpubs.stanford.edu:8090/778/1/2006-13.pdf
> 3. kmeans|| : http://theory.stanford.edu/~sergei/papers/vldb12-kmpar.pdf
> For very large data sets, or for large values of k, the kmeans|| method is
> preferred as it provides the same approximation guarantees as kmeans++ and
> requires lesser number of passes over the input data.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
