Revision: 17312
          http://sourceforge.net/p/gate/code/17312
Author:   valyt
Date:     2014-02-15 12:59:04 +0000 (Sat, 15 Feb 2014)
Log Message:
-----------
Slightly better comment for whoever gets lost in these parts in the future.

Modified Paths:
--------------
    
gate/trunk/plugins/Text_Categorization/src/gate/ml/categorization/TextCategorizationPR.java

Modified: 
gate/trunk/plugins/Text_Categorization/src/gate/ml/categorization/TextCategorizationPR.java
===================================================================
--- 
gate/trunk/plugins/Text_Categorization/src/gate/ml/categorization/TextCategorizationPR.java
 2014-02-14 18:49:50 UTC (rev 17311)
+++ 
gate/trunk/plugins/Text_Categorization/src/gate/ml/categorization/TextCategorizationPR.java
 2014-02-15 12:59:04 UTC (rev 17312)
@@ -397,13 +397,17 @@
         double probability = 1.0;
         // do we need to check probabilities?
         if(categoryThresholds != null && categoryThresholds[label] > 0) {
-          // prob[i] is the value of $\theta_T x$ 
+          // prob[i] is the value of $\theta^T x$ 
           // Once mapped through the logistic function, this becomes the 
           // probability of the instance belonging to class i, as opposed to 
           // all other classes.
-          // LIBLINEAR would normalize this so that all probs sum up to 1,
-          // but we don't want that, as we're only interested in the confidence
-          // the model has in this particular classification.
+          // LIBLINEAR normalizes the values so that all probabilities sum up 
+          // to 1 (i.e. the output is a probability distribution over the set 
+          // of classes). They should really be using softmax for that, but 
+          // anyway... 
+          // However, we don't want that, as we're only interested in the 
+          // confidence the model has in this particular classification, and 
we 
+          // don't know for sure that the classes are mutually-exclusive 
           probability = probs[label];
           // convert to an actual probability, by applying the logistic 
function
           probability = 1 / (1 + Math.exp(-probability));

This was sent by the SourceForge.net collaborative development platform, the 
world's largest Open Source development site.


------------------------------------------------------------------------------
Android apps run on BlackBerry 10
Introducing the new BlackBerry 10.2.1 Runtime for Android apps.
Now with support for Jelly Bean, Bluetooth, Mapview and more.
Get your Android app in front of a whole new audience.  Start now.
http://pubads.g.doubleclick.net/gampad/clk?id=124407151&iu=/4140/ostg.clktrk
_______________________________________________
GATE-cvs mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/gate-cvs

Reply via email to