You are viewing a plain text version of this content. The canonical link for it is here.

Posted to dev@mahout.apache.org by Grant Ingersoll <gs...@apache.org> on 2011/11/01 19:20:57 UTC

Train/TestNewsGroups with SGD

I'm working on https://issues.apache.org/jira/browse/MAHOUT-857.  Each time I run it, I get different answers for SGD for the confusion matrix, which is  presumably due to the randomness  built in.  However, is there a way to set the seed so one can reproduce results for actually testing the code?  

Thanks,
Grant

Re: Train/TestNewsGroups with SGD

Posted by Grant Ingersoll <gs...@apache.org>.

On Nov 1, 2011, at 2:45 PM, Sean Owen wrote:

> RandomUtils.setTestSeed() (or something like that) makes all the RNGs
> deterministic -- well if they are using RandomUtils.

I see it in use in at least one place.

> 
> On Tue, Nov 1, 2011 at 6:20 PM, Grant Ingersoll <gs...@apache.org> wrote:
>> I'm working on https://issues.apache.org/jira/browse/MAHOUT-857.  Each time I run it, I get different answers for SGD for the confusion matrix, which is  presumably due to the randomness  built in.  However, is there a way to set the seed so one can reproduce results for actually testing the code?
>> 
>> Thanks,
>> Grant

Re: Train/TestNewsGroups with SGD

Posted by Sean Owen <sr...@gmail.com>.

RandomUtils.setTestSeed() (or something like that) makes all the RNGs
deterministic -- well if they are using RandomUtils.

On Tue, Nov 1, 2011 at 6:20 PM, Grant Ingersoll <gs...@apache.org> wrote:
> I'm working on https://issues.apache.org/jira/browse/MAHOUT-857.  Each time I run it, I get different answers for SGD for the confusion matrix, which is  presumably due to the randomness  built in.  However, is there a way to set the seed so one can reproduce results for actually testing the code?
>
> Thanks,
> Grant