Summary of Methods
•
Simple
: choose unlabeled instance closest to
current SVM
•
•
Ratio
: choose unlabeled instance for which
m
+
and
m
-
are similar in size
•