I reverted to the list form for the paper layout.  Aesthetically, I
like encoding a list in a list rather then a list in a paragraph and
space does not seem to be an issue.  There is some sort of compact
list thing which is possible in latex as well if space becomes an
issue.

>In 2.2: Is "decided" correct? I'd write "certain". Didnt change it though.

I like "decided" because we are making a "decision" by weighted
voting.  But my inclination is not strong either way.

footnote 6 rewritten

Altered hoeffding bound statement slightly.

Fiddled with the 'example' section in various ways.

I do not understand the comment: "In all applications of MRED we know
of, the class of discriminant functions {\em is} uniformly bounded if
we restrict ourselves to a bounded region in ${\cal X}$ where all the
data lies with high probability."  Presumably, you mean something like
"if you use an exponential family with compact support then the
probability is always greater then zero so the discriminant is
bounded."  This seems ... like a bit of a stretch.  It _is_ bounded.
But the bound is exponentially large, and given that the variance
increases with the bound^2, this will not be a very strong statement
in practice.  I think I would prefer to not make this statement
because it is sort of misleading.  Removing and adding a note about
the algorithm directly motivated by our bound.  I like the note
because it is interesting to think about the algorithm directly
motivated by the bound _and_ it helps emphasize how the MRED criteria
doesn't quite fit.