Title:        DNF is Properly PAC-learnable in Polynomial Time
Authors:      R. Bungler and T. Mangler
Email:        rbungler@cs.unowhere.edu, tmangler@cs.unowhere.edu
Home_page:    http://www.cs.unowhere.edu/~rbungler,
              http://www.cs.unowhere.edu/~tmangler
Affiliation:  University of Nowheresville
Address:      Computer Science Department,
              1234 Dusty Death Road, 
              Nowhersville, BS 56789-0123
Thanks_to:    This research was supported by grant XYZ-456789
              from the National Signs Foundation.
Abstract:
         We show that boolean functions expressed as
disjunctive normal form formulas are properly PAC-learnable
in the model of Valiant. Not only does our learning algorithm
not use any kind of queries in the learning process, it also 
does no sampling of the target function. Instead, we use
a new technique---Divine Inspiration---to show our results.
We believe that our ideas have immediate and far-reaching 
impractical applications beyond learning. Some of these will
be discussed in a forthcoming paper.