You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@opennlp.apache.org by James Kosin <ja...@gmail.com> on 2011/08/28 17:59:34 UTC

testMaxentOnPrepAttachDataWithParamsDefault FAILURE

Any reason why this test would fail?
I'm getting a difference of:

Accuracy on PPA devset: (3266/4039) 0.8086159940579352

Expected:   0.7997028967566229
Actual:        0.8086159940579352

James

Re: testMaxentOnPrepAttachDataWithParamsDefault FAILURE

Posted by James Kosin <ja...@gmail.com>.
On 8/28/2011 5:09 PM, Jörn Kottmann wrote:
> On 8/28/11 11:06 PM, James Kosin wrote:
>> Thanks, I thought the cutoff was always defaulting to 5.
>
> Yes it does, but I got a little confused when I did it because there 
> is a "second"
> cutoff handling (usually not used) in the GISTrainer class, and I 
> mistakenly though
> I am using it.
>
> Jörn
>
No problems then.  I just thought for a moment there there was something 
wrong with Java.

Re: testMaxentOnPrepAttachDataWithParamsDefault FAILURE

Posted by Jörn Kottmann <ko...@gmail.com>.
On 8/28/11 11:06 PM, James Kosin wrote:
> Thanks, I thought the cutoff was always defaulting to 5.

Yes it does, but I got a little confused when I did it because there is 
a "second"
cutoff handling (usually not used) in the GISTrainer class, and I 
mistakenly though
I am using it.

Jörn

Re: testMaxentOnPrepAttachDataWithParamsDefault FAILURE

Posted by James Kosin <ja...@gmail.com>.
On 8/28/2011 1:11 PM, Jörn Kottmann wrote:
> On 8/28/11 6:58 PM, Jörn Kottmann wrote:
>> I did a change under the assumption that cutoff == 1 is equal
>> to not setting the cutoff at all, I will check to what it is set instead
>> now. 
>
> Sure, cutoff is 5 in this case. That explains the difference.
>
> Jörn
Thanks, I thought the cutoff was always defaulting to 5.

James

Re: testMaxentOnPrepAttachDataWithParamsDefault FAILURE

Posted by Jörn Kottmann <ko...@gmail.com>.
On 8/28/11 6:58 PM, Jörn Kottmann wrote:
> I did a change under the assumption that cutoff == 1 is equal
> to not setting the cutoff at all, I will check to what it is set instead
> now. 

Sure, cutoff is 5 in this case. That explains the difference.

Jörn

Re: testMaxentOnPrepAttachDataWithParamsDefault FAILURE

Posted by Jörn Kottmann <ko...@gmail.com>.
On 8/28/11 5:59 PM, James Kosin wrote:
> Any reason why this test would fail?
> I'm getting a difference of:
>
> Accuracy on PPA devset: (3266/4039) 0.8086159940579352
>
> Expected:   0.7997028967566229
> Actual:        0.8086159940579352

I did a change under the assumption that cutoff == 1 is equal
to not setting the cutoff at all, I will check to what it is set instead
now.

Jörn