mahout-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Yutaka Mandai <20525entrad...@gmail.com>
Subject Re: Speed up LDA in Mahit 0.9
Date Fri, 08 May 2015 14:15:32 GMT
If it's small enough to fit in memory, setting MAHOUT_LOCAL="TRUE" should drive you crazy!

I've suffered a lot from running LDA(CVB0) on even on EMR. If you believe your data is small
enough, then the local is the best.

Regards,,,
Y.Mandai

iPhoneから送信

2015/05/07 20:12、mw <mw@plista.com> のメッセージ:

> As far as I understood, the runtime complexity is O(N*T*D),
> where N is the number of words, T the number of topics and D the number of documents.
> 
> So you can try e.g. to reduce the number of words.
> 
>> On 05/05/2015 10:36 AM, Donni Khan wrote:
>> Hello Mahout Users,
>> 
>> I'm runing LDA job (Mahout 0.9) by using java code, but to run the
>> algorithm on the small dataset is taking much time.
>> Is there any way to speed up the prcessing time (like changing the
>> parameter values)?
>> 
>> Thanks in advance,
>> Donni
> 

Mime
View raw message