In this method the value of an attribute is compressed into a binary
word of width with a fixed number of
bits set to
. This
number is called the weight. The problem is how to represent
variables so that they can match with anything. In COLOMB
three possibilities to do this are proposed.
The main advantage of this method is that one can currently
construct hardware that
handles up to clauses and more in the presented manner. Together
with the linear searching hash-function one reaches a very high efficiency.
Another key property is that m-in-n-coding results in highly compressed
code,
so that large clause code can be stored separately (externally) from the
small index code and only single rules are loaded.