W.S Havens. Intelligent Backtracking in the Echidna Constraint Logic Programming System. Technical Report CSS-IS 92-12, Simon Fraser University, 1992.  Home/Search   Document Details and Download   Summary   Related Articles   Check

.... accuracy for the sake of faster execution of programs [Bru78, PP79, Bru80, Bru81, PP80b, PP80a, BP84, CCF88, Dra88, CS91, Mal90] Some other methods go even further and abandon failure analysis altogether accepting more moderate conflict computation methods based on analysis of data dependencies [KL88, MTK89, Hav92]. The advantage of such methods is the low overhead during forward execution of logic programs. Their main disadvantage is the loss of accuracy of backtracking information which sometimes results to more inefficient search than naive backtracking even for problems which are considered very good ....

....of computation is different and so are the failure analysis methods. Constraint Logic Programming systems over different domains need different failure analysis techniques. FFS93] has developed an intelligent backtracking version of a Constraint Logic Programming language over finite domains. [Hav92] has developed an intelligent backtracking method for a Constraint Logic Programming system over real intervals. DBB91] was the first theoretical study of failure analysis and intelligent backtracking in CLP( languages and [BLH94] implements a version of intelligent backtracking in a CLP( ....

[Article contains additional citation context not shown here]

W.S Havens. Intelligent Backtracking in the Echidna Constraint Logic Programming System. Technical Report CSS-IS 92-12, Simon Fraser University, 1992.

....considered. Therefore, some intelligent backtracking schemes sacrifice failure analysis accuracy for the sake of faster execution of the programs [CCF88, CS91, PP80b, PP80a] Some other methods go even further and abandon failure analysis altogether, accepting more moderate data dependency methods [KL88, Mal90, Hav92]. As it will be shown later this is not an appropriate approach in Constraint Logic Programming especially when the network of constraints is dense. Recently, as reported in [CS91] intelligent backtracking has been successfuly integrated into the WAM machinery [AK92] The results are very ....

....subsets of constraints on failure. This algorithm must operate on the same constraint representation which the solver uses during the derivation sequence of a CLP system. ffl and finally compare the failure analysis methods proposed here to the two other methods proposed in the literature [DBB91, Hav92] for the domain of linear constraints over the reals. The rest of this paper is organised as follows. In section 2 the motivations for using intelligent backtracking in a CLP( system are briefly presented. Section 3 illustrates some of the issues a failure analysis procedure must handle, by ....

[Article contains additional citation context not shown here]

William S. Havens. Intelligent Backtracking in the Echidna Constraint Logic Programming System. Technical Report CSS-IS 9212, Simon Fraser University, 1992.

....like AC [10] AC 4 [11] and AC 5 [17] for finite domains, or unification over first order terms are instances of such local solvers. Up to now, the deletion of unification constraints in Prolog like systems has been tackled by so called selective reset in extended intelligent backtracking methods [3, 7]. However, even if such methods can be extended to finite domain constraints [5, 14] they rely on complex dependency recording mechanisms that amount to runtime overhead, even if no deletion is later used. Instead, we would like to take advantage of the dependency information already present in ....

W. S. Havens. Intelligent Backtracking in the Echidna Constraint Logic Programming System. Research Rep. CSS-IS TR 92-12, Simon Fraser University, Vancouver, Canada, 1992.

....untouched. It is important to note that we do not want a justification based method that would compute and keep explicitely the dependencies between constraints resulting from consistency checking and domain reduction in a TMS [7] or ATMS style [6] This has been proposed in CLP by [3] and [9] and developed in [8] in the Concurrent Constraint framework by [4] in the CSP framework by [17, 12] and for Hierarchical CLP in [14] However all those methods have the drawback to slow down the constraint propagation while computing the dependencies and are therefore costly when no ....

W. S. Havens. Intelligent Backtracking in the Echidna Constraint Logic Programming System. Research Rep. CSS-IS TR 92-12, Simon Fraser University, Vancouver, Canada, 1992.

....propagation techniques (e.g. AC [19] AC 4 [20] AC 5 [28] for finite domains, or unification over first order terms. Up to now, the deletion of unification constraints in Prolog like systems has been tackled by so called selective reset in extended intelligent backtracking methods, see [3] or [13]. However, even if such methods can be extended to finite domain constraints [6, 25] they rely on complex dependency recording mechanisms that amount to runtime overhead, even if no deletion is later used. To avoid these problems, our idea is to take advantage of the dependency information ....

W. S. Havens. Intelligent Backtracking in the Echidna Constraint Logic Programming System. Research Rep. CSS-IS TR 92-12, Simon Fraser University, Vancouver, Canada, 1992.

....of committing to user choices. We show how standard backtracking algorithms like chronological backtracking and conflict directed backtracking do not commit to user choices. Unnecessarily undoing user choices has previously been recognized as a problem with standard backtracking algorithms. Havens [6] calls this situation backtracking the user and proposes an algorithm that helps to avoid the problem. This algorithm is described in section 4 where we show that the algorithm does not commit to user choices in all cases. Finally, we present a modification of Ginsberg s dynamic backtracking ....

....variable or user variable. For the above example, we build a partial solution (x; 1) y; 1) but then fail to find a value than can be assigned to the variable a. The assignment of y to 1 is undone even though there is a solution with y assigned to 1. Methods for intelligent backtracking [2, 3, 6, 9] are capable of jumping back to a culprit that is not the most recent assignment. For example, using conflict directed backjumping [9] the partial solution (x; 1) y; 1) would be built but the failure at the variable a is recognized as being independent of the assignment of y. The failure is ....

[Article contains additional citation context not shown here]

William S. Havens. Intelligent backtracking in the echidna constraint logic programming system. International Journal of Expert Systems: Research and Applications, 5(4):319--343, 1992.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC