BUG REPORT

Author: Alex Klein

Problem: Training Classifier twice without changing tabs causes CEBL to hang until 
Stop Training button is pressed or tab is switched.

Error: The training thread was not terminated correctly after successfully training, 
preventing it from responding to the new training request.

Solution:  TabTraining.timedUpdateInterface:  added command to stop the training thread 
upon successful completion of training.

__________________________________________________________________________________________

Author: Alex Klein

Problem: Cebl hangs when switching between data sources

Status: I have been unable to reproduce this bug.  I suspect that the cause was the data 
source being changed while the training thread was still active for another data source.  
Since the training thread bug has been fixed, I think this will no longer be an issue, but 
I have been unable to confirm this.

__________________________________________________________________________________________

Author: Alex Klein

Problem: Classification bar shoots into the top left corner of window.

Error: In QDA.cpp conditions arise that cause Not a Number (-nan) entries into matricies, 
eventually causing proportions in TabRealTimeClassification.timedUpdateInterface (vector 
std_props) to have -nan entries.

Status: I attempted to check for the -nan entries in std_props and set the proportions to 
1.0/numClasses (even split) using

	if(!(std_props[0] >= 0 || std_props[0] < 0))

This worked in that it prevented the bar from shooting into the top left corner, but did 
not prevent future -nan entries, and simply caused it to appear as if the classifier was 
constantly evenly rating them.  I did not commit this "fix".  Real problem lies in the 
QDA algorithm.
