[prev in list] [next in list] [prev in thread] [next in thread] 

List:       wekalist
Subject:    Re: [Wekalist] Neural Network soft output
From:       Eibe Frank <eibe () cs ! waikato ! ac ! nz>
Date:       2003-12-16 23:58:05
[Download RAW message or body]

Hi Dave,

Do you want to get at those values from within a program? Then you can 
use the distributionForInstance() method. At the command-line, you can 
use the -p option. In the Explorer you can select "Output text 
predictions on test set" under "More options..." once you have selected 
a test set that you want to get the predictions for.

Cheers,
Eibe

On Wednesday, December 17, 2003, at 12:23  PM, 
wekalist-request@list.scms.waikato.ac.nz wrote:

> From: "Dave Bazell" <bazell@comcast.net>
> Date: Tue Dec 16, 2003  1:30:56  PM Pacific/Auckland
> To: <wekalist@list.scms.waikato.ac.nz>
> Subject: [Wekalist] Neural Network soft output
> Reply-To: Dave Bazell <bazell@comcast.net>
>
>
> I have asked this twice before but I got no reply.  Can I get soft 
> outputs
> from the backprop neural network in WEKA?  If I have three output 
> classes I
> want to know the activation for each output node, not just the assigned
> class.  Thus, I would like an output that says:
>
> 0.3455, .5910, 0.0635
>
> rather than just telling me that the example was assigned to class # 
> 2.  The
> outputs don't have to be normalized to sum to 1.00, I can do that.
>
> Can someone answer this for me?
>
> Thanks,
>
> Dave


_______________________________________________
Wekalist mailing list
Wekalist@list.scms.waikato.ac.nz
https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
[prev in list] [next in list] [prev in thread] [next in thread]