Services & ResourcesWolfram ForumsMathGroup Archive

[Date Index] [Thread Index] [Author Index]

Re: Clustering algorithm Mathematica 7

• To: mathgroup at smc.vnet.net
• Subject: [mg129195] Re: Clustering algorithm Mathematica 7
• From: Matthias Odisio <matthias at wolfram.com>
• Date: Thu, 20 Dec 2012 03:25:34 -0500 (EST)
• Delivered-to: l-mathgroup@mail-archive0.wolfram.com
• Delivered-to: l-mathgroup@wolfram.com
• Delivered-to: mathgroup-newout@smc.vnet.net
• Delivered-to: mathgroup-newsend@smc.vnet.net
• References: <200911201139.GAA03443@smc.vnet.net> <he895k$q7k$1@smc.vnet.net> <20121216060625.83C4D6906@smc.vnet.net>

Hello,

On 12/16/12 12:06 AM, martin.lottner at gmail.com wrote:> Hi,
>
> I have started using the clustering methods in Mathematica 9 and
> I'd like to know whether the "FindClusters" function with the
> "optimization" method still uses k medoids in Version 9.

Yes it does.

> Could you please specify on which methods/publications the initial
> seed choice is based? (The documentation states that the algorithm
> "starts by building a set of k representative objects")
> What lies behind the "KMeans" option in ClusterComponents? Is this
> just an implementation of the naked k means algorithm with a
> random choice of initial seeds or is there some seed selection
> process involved as a first step?

I would think it's a random process. You can control it using the
"RandomSeed" (sub-) option:

     FindClusters[l, n, Method -> {"Optimize", "RandomSeed" -> s}]
     ClusteringComponents[l, n, "RandomSeed" -> s]


Would the bottom half of this tutorial provide more helpful information?
http://reference.wolfram.com/mathematica/tutorial/PartitioningDataIntoClusters.html


Matthias Odisio
Wolfram Research

• References:
  • Re: Clustering algorithm Mathematica 7
    • From: martin.lottner@gmail.com