DM: learning m-of-n concepts

[Patrick & Linda Murphy]

On Tue, 29 Jul 1997 duttond@bre.co.uk wrote:

> Since completing a PhD (looking at efficient induction of decision 
>trees, e.g.
> for nasty things like m-of-n problems) Ive been working at the 
>Building

Dave,

I'd be interested in what research you've done involving the 
induction of
m-of-n concepts using decision trees.  I produced an algorothm a few 
years
back called ID2-of-3 that induces m-of-n concepts as discriminators in
decision trees.  Given the popularity of Quinlan's C4.5 decision tree
induction algorithm, I've been thinking that it's about time to pull
ID2-of-3 "out of moth balls" and present it to the datamining 
community as
a viable alternative to inducing trees, and more importantly 
extracting
rules, that only use simple feature-value tests.  Since m-of-n 
concepts
are satisfied  by a sufficient summing of evidence, they may be a more
appropriate bias for real world data.

If you could provide me with a list of your m-of-n related references
as well as any others that you've come across, I'd be very 
appreciative.

Thanks in advance,

- Patrick