mahout-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jeff Eastman <j...@windwardsolutions.com>
Subject Re: Un-observing in a Canopy
Date Sun, 04 Mar 2012 21:29:52 GMT
You can cause a canopy to un-observe a vector by observing it with a -1 
weight. This will have the effect of subtracting all influence of the 
vector on observations of that canopy. But you can only do that before 
you computeParameters, as all observation semantics is reset by that 
operation and converted to center and radius (mean & std).  But hmn, it 
might also work after computeParameters if you observe it with a weight 
of  -1/numObservations and computeParameters again. Double check the 
math on this.

On 3/4/12 9:14 AM, Paritosh Ranjan wrote:
> There is a way to recalculate the parameters, as I already mentioned
>
> canopy.computeParameters(),
>
>
> but I can not find any way to un-observe a Canopy.
>
> On 04-03-2012 19:58, Benson Margulies wrote:
>> On Sun, Mar 4, 2012 at 9:10 AM, Paritosh Ranjan<pranjan@xebia.com>  
>> wrote:
>>> Not sure whether I got your question correctly.
>>> Are you looking for canopy.computeParameters().
>> 'observe' adds an item to the cluster for which Canopy is calculating
>> parameters when asked.
>>
>> I'm just confirming that there's no way to 'subtract' an item and
>> recalculate the parameters.
>>
>>
>>>
>>> On 04-03-2012 05:09, Benson Margulies wrote:
>>>> So, I'm messing around with a clustering problem where I want to move
>>>> items from cluster to cluster and maintain centroids. Lazy person that
>>>> I am, I am using the Canopy class. But it looks to me like I have to
>>>> recreate the Canopy when I pull something out. Have I missed anything?
>>>
>
>
>


Mime
  • Unnamed multipart/mixed (inline, None, 0 bytes)
View raw message