 
  
  
  
  
 Next: Question 6
Up: Evaluating Hypothesis and Experimental 
 Previous: Problems with McNemar
 
-  Given a learning algorithm   and a large dataset and a large dataset , what
is the accuracy of the classifiers produced by , what
is the accuracy of the classifiers produced by when trained on
new training sets of specified size when trained on
new training sets of specified size ? ?
-  Has not received much attention in the literature.
-  The DELVE approach is to subdivide   into a test set and
several disjoint training sets of size into a test set and
several disjoint training sets of size . . is trained on each
of the training sets and the resulting classifiers are all tested on
the test set.  The average performance on the test set estimates the
accuracy on new runs. is trained on each
of the training sets and the resulting classifiers are all tested on
the test set.  The average performance on the test set estimates the
accuracy on new runs.
 
Patricia Riddle 
Fri May 15 13:00:36 NZST 1998