Hi all,
Can anyone help me out with understanding optimization of my Rete
network. "Chapter 10. The Rete Algorithm" briefly discusses the issues,
and the second last paragraph is particularly tantalizing, but I'm
unclear on some of the details:
- Using (watch compilations) I have noticed in my fact/rule base that
there are many reuses of 1 input nodes (i.e. =1 on the output), but
almost no sharing of 2 input nodes (i.e. almost no =2). This surprises
me since many of the rules have similar but not quite the same patterns
on the LHS.
- What are the 'rules' in the algorithm for 2 input nodes to be shared?
Can you describe this in terms of pattern matching components?
- Is the node sharing algorithm sensitive to the order of facts and/or
fact slots on the LHS?
- There are a number of (not ...) patterns in the rulebase. Does 'not'
logic have special implications?
- Would the use of unordered facts over ordered facts substantially
improve efficiency?
I notice that ordered facts require a 'count fields' node (most
obvious from the 'view' GUI). This seems like an O(F) impact at worst -
correct?
(side note: is there any way to scroll right on the 'view' display...)
Thanks to anyone who feels like tackling this...
Jack
--
* This email is packaged by intellectual weight, not volume. Some
settling of contents may occur during transmission.
---------------------------------------------------------------------
To unsubscribe, send the words 'unsubscribe jess-users [EMAIL PROTECTED]'
in the BODY of a message to [EMAIL PROTECTED], NOT to the
list (use your own address!) List problems? Notify [EMAIL PROTECTED]
---------------------------------------------------------------------
