Re: [eclipse-clp-users] strange behaviour of minimize/bb_min

From: Stephan Schiffel <stephan.schiffel_at_...69...>
Date: Thu, 5 Aug 2010 08:52:07 +0200
Hi all,

On Thursday 05 August 2010, Joachim Schimpf wrote:
> > Stephan Schiffel wrote:
> >> ?- hash_create(H), minimize((member(X,[3,2,1]), hash_add(H,solution,X)),
> >> X). Found a solution with cost 3
> >> Found a solution with cost 2
> >> Found a solution with cost 1
> >> Found no solution with cost -1.0Inf .. 0
> >>
> >> H = hash(4, 0, [])
> >> X = 1
>
> To get the behaviour you want, use a store that survives backtracking
> (record, bag, store, shelf), e.g.
>
> ?- record_create(H),
>    minimize((member(X, [3, 2, 1]), recordz(H, X)), X),
>    recorded_list(H, Xs).

Thanks for the suggestion, but this is not what I want. I really just want to 
remember some information about the best solution, not about all solutions.
The output I'd like would be:

? hash_create(H), minimize((member(X,[3,2,1,4,5]), hash_add(H,solution,X)), 
X).
...
H = hash(4, 1, [solution -> 1])
X = 1

That means as long as I'm not using any non-logical storage, I'd like 
minimize(Goal, Cost) to behave exactly as calling Goal and selecting 
the "right" branches on choice points instead of the first one first. So far, 
I thought of lib(hash) as logical storage, because it behaves well on 
backtracking. I know that it is not strictly logical, because it uses 
setarg/3 internally.

Kish's e-mail explains the problem. Unfortunately, that means that using 
hash-tables is not a good idea because they don't behave well, at least not 
together with lib(branch_and_bound).

I wonder if there is a way to implement hash-tables efficiently in a logical 
way, i.e., without using setarg/3. The problem could be easily solved with 
something like setarg(+N, +OldTerm, ?Arg, ?NewTerm), that does not change 
OldTerm but constructs a new one instead. However, avoiding the construction 
of a new term is probably what makes setarg/3 so efficient, right?

Regards,
Stephan
-- 
                              Stephan Schiffel
                       Technische Universität Dresden,
                 Fakultät Informatik, 01062 Dresden, Germany
           Tel: [49] (351) 463-39243    Fax: [49] (351) 463-37924
                  Email: stephan.schiffel_at_...69...
Received on Thu Aug 05 2010 - 06:52:21 CEST

This archive was generated by hypermail 2.3.0 : Tue Apr 16 2024 - 09:13:20 CEST