Andrew suggests we look into LR and, more specifically, l1 LR, which is
a technique that does something like:

  x1 + x2 + x1x2 + ...

The l1 modification of LR generates useful meta-like rules that should
not be too overoptimized for SA.

So, we should try both.  :-)

Daniel

-- 
Daniel Quinlan
http://www.pathname.com/~quinlan/

Reply via email to