Puget, J-F. & Leconte, M. (1995) Beyond the glass box: constraints as objects. Proc. of International Symposium on Logic Programming (ILPS'95), MIT Press.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....low level operations, which are difficult to program efficiently in this setting. The indexical approach [19, 3] allows the user to program some new constraints. But it has no support to apply more sophisticated algorithmic techniques to implement new constraints (see also Section 8 and [13]) In Jan Mal uszynski, editor, Proceedings of the International Logic Programming Symposium, pages 149 163. The MIT Press, 1997. On the other hand, combinatorial problems can be tackled in a language like together with a dedicated library for constraint solving (see e.g. ILOG [8] Although, ....

....by constructing them with indexicals. Indexicals are terms x in r where r defines how x is constrained and on what event an indexical is resumed. The indexicals which realize a single constraint exist independently of each other, i.e. the constraint is not available first class (see also [13]) Thus, algorithmic techniques employing global reasoning on all arguments of the constraint cannot be incorporated in this setting. AKL(FD) 2] implements indexicals in a concurrent constraint setting where local computation spaces are employed in so called deep guards. Hence, there are ....

J.-F. Puget and M. Leconte. Beyond the glass box: constraints as objects. In Proceedings of the International Symposium on Logic Programming, pages 513--527, 1995.

....information of the constraints. For the alldifferent constraint at least five di#erent filtering algorithms exist, each achieving a di#erent kind of consistency, or achieving it faster [28, 20, 26, 23] Although the constraint is mentioned more or less deeply in a variety of papers and books (e.g. [27, 15, 22]) to our knowledge there is no work that collects all e#ort that has been put into this constraint. This paper therefore tries to give an overview of the alldifferent constraint, which will be outlined in the next section. 1.3. Outline of the paper This paper is organized as follows. First we ....

J.-F. Puget and M. Leconte. Beyond the glass box: Constraints as objects. In International Logic Programming Symposium, pages 513--527, 1995.

.... of Ordinary Di erential Equation (ODE) constraints [16, 18] but have yet to incorporate the higher level ODE contractors of Deville, Janssen, and van Hentenryck[10] or the multidimensional Taylor approximations of Berz and Ho st atter [5] Our metalevel approach is similar to Puget and Leconte s [36], except that we manipulate constraints at the metalevel while they advocate giving constraints rst class status in the underlying constraint solver. Another approach to mixing RISC CISC style constraint solving is the DECLIC system[13] which di ers from ours in that it provides both RISC and ....

J.-F. Puget and M. Leconte. Beyond the Glass Box: Constraints as Objects. Proceedings of International Logic Programming Symposium, 1995.

....problems that can then be solved using state of the art solvers. These modeling languages do not require speci c programming skills and can be used by a wide audience. Constraint programming languages such as CHIP [6] Prolog III [4] Eclipse [7] Prolog IV [5] Oz [16, 17] Salsa [11] and [14] have orthogonal strengths. Their constraint vocabulary and their underlying solvers go beyond traditional linear and nonlinear constraints and support logical, higher order, and global constraints. They also make it possible to program search procedures to specify how to explore the search space. ....

J-F. Puget and M. Leconte. Beyond the Glass Box: Constraints as Objects. In Proceedings of the International Symposium on Logic Programming (ILPS-95), Portland, OR, November 1995.

....Unfortunately, CHIP [22] and Prolog IV [18] could not be included since they were not available to us. Although the constraints systems have already been tested on a number of traditional benchmarks, most comparative work has been done by the language implementers themselves ( 3] 5] 8] 17] [19]) Furthermore, these benchmarks have been used in the development as well as the assessment of the languages so that such tests 1 are biased. Neither of the authors of this paper is a designer of any of the CLP languages studied here so that the comparison is more impartial than previous ones. ....

PUGET J-F. AND LECONTE M., Beyond the Glass Box: Constraints as Objects. In Proc. of ILPS'95, Portland, December, 1995.

....algorithms (for instance, hull consistency or box consistency according to the number of occurrences of a variable in a constraint) and no computation of individual fixed points of the techniques to be combined. In other words, a solver must not be taken as a black box, but as a glass box [22] where the elementary components can be extracted in order to be re used by more general algorithms. The research reported in this paper may be extended in many ways. First, the accuracy of narrowing algorithms may be improved through the use of several constraint interval extensions. ....

J.-F. Puget and M. Leconte. Beyond the Glass Box: Constraints as Objects. In Proc. Int. Logic Programming Symposium, 1995.

....in terms of integration and modularity (Puget, 1994) While the indexical approach (Carlson, Carlsson, Diaz, 1994) provides 1 Thereby referred as constraint programming systems. 2 a glass boxapproach to program new constraints, it is insufficient as constraints cannot be manipulated directly (Puget Leconte, 1995). Another problem with Prolog is that it uses depth first search for resolution or searching. It is unfortunate because wecannotintegrate specialized search strategy to solve differenttypes of problems (Schulte, 1997b) Oz, with its space abstraction (Schulte, Smolka, Wurtz, 1994) is thusavery ....

....with related works and suggest directions for future research. 8.1 Major Contributions The recent constraint programming systems are no longer tied to a logic programming framework. They offer us more flexibility of extending their constraint programming systems by programming new propagators (Puget Leconte, 1995# Muller Wurtz, 1997) and search algorithms (Caseau, 1994# Caseau Laburthe, 1996# Schulte, 1997b) Unfortunately, these existing systems originated from a Prolog framework and havethus retained certain characteristics from Prolog. For example, in the Ilog Solver, the solver searches for a ....

Puget, J.-F., & Leconte, M. (1995). Beyond the Glass Box: Constraints as Objects. In Proceedings of the International Symposium on Logic Programming (pp. 513--527).

....by lowerlevel constraints would include inequality constraints for all possible variable pairs of the set V. But the application of the global constraint with a specific data representation and satisfaction methods can yield much better performance (e.g. by a demon observed array representation [PL95]) The notion of global constraints can support local search approaches as well. It is transferred to a local search context in the following subsections. 2.1. Global Constraints from a Local Search Perspective. The central issue in local search is the transition from one state to the successor ....

....worsened or failed to improve the search behavior. The only useful technique was randomization, though this was true for some decision points only, the advantage decreasing with the inclusion of stronger knowledge. The concept of global constraints was originally used for refinement search (e.g. [LeP94, PL95]) Transferring it to a local search context makes it possible to get an efficient and declarative handle on local search, while preserving features like reusability and maintenance. Further work will include concepts to temporally focus the search on specific aspects and the integration of ....

Puget, J.-F., and Leconte, M. 1995. Beyond the Glass Box: Constraints as Objects. In Proceedings of the 1995 International Logic Programming Symposium (ILPS'95), 513--527.

....to handle this sequence constraint efficiently. This constraint is typically used for complex timetable problems to specify regulations rules (e.g. any employee has at least twice a month a two day rest) Another kind of sequence constraint is the Global Sequencing Constraint [15] of IlogSolver [13]. This constraint is used to specify the number of successive items having their values in a given set. This constraint is a generalization of the global cardinality constraint [16] and is filtered by the same method. Our problem is different because we need to constrain not only the value of ....

Puget, J.-F. and Leconte, M. "Beyond the Glass Box: Constraints as Objects," ILPS'95, Portland, Oregon, 1995.

....Instead of having one single solver for linear constraints, it can be useful to allow the use of many of them, each of which being dedicated to solving a subproblem. The C library Ilog Solver provides an object oriented framework for designing and implementing efficiently complex constraints [PL95]. Our prototype linear constraint solver was therefore implemented in Ilog Solver as a global constraint over variables whose domain ranges over floating point numbers . It is important to notice that this constraint has several peculiarities, the main of which being that is can be incrementally ....

J-F. Puget and M. Leconte. Beyond the Glass Box: Constraints as Objects. Proceedings of the International Logic Programming Symposium, Portland, December 1995.

....specific methods. This approach to programming was successfully realized in a number of programming languages, notably constraint logic programming languages. Up to now, the most successful approach to imperative constraint programming is the object oriented approach taken by ILOG Solver (see [16], 9] In this system constraints and variables are treated as objects and are defined within a C class library. Thanks to the class encapsulation mechanism and the operator overloading capability of C , the user can see constraints almost as if they were a part of the language. A similar ....

J.-F. Puget and M. Leconte. Beyond the glass box: Constraints as objects. In Proc. of the 1995 International Symposium on Logic Programming, pages 513--527, 1995.

....the given constraints. General purpose search mechanisms can master smaller tasks, but realworld problems are often out of reach. A way to approach larger and more complex problems is to exploit domain specific information. This can be achieved by the application of so called global constraints [PL95], instead of using a standard problem encoding with simple primitives such as linear inequalities and the like. A global constraint is a replacement for a set of lower level constraints, where additional domain knowledge allows the application of specialized data representations and algorithms to ....

....constraints would include inequality constraints for all possible variable pairs of the set V. But the application of the global constraint with a specific data representation and satisfaction methods can yield a much better performance (e.g. by a demon observed array representation, see [PL95]) Global constraints are a kind of compromise between generality and efficiency (see fig. 1) They can efficiently master complex relations, but should be fine grained enough to be applicable in a range of different problems. Global constraints do not have to cover disjunct problem areas, as ....

Puget, J.-F., and Leconte, M. 1995. Beyond the Glass Box: Constraints as Objects. In Proceedings of the 1995 International Logic Programming Symposium (ILPS'95), 513--527.

....1 ; Xn g, where each variable is associated with a domain D 1 ; Dn , and a set of constraints C = fC 1 ; Cm g over these variables. Constraints are relations between variables, that restrict the possible value assignments. The Excalibur model applies so called global constraints (Puget Leconte 1995), instead of using a standard problem encoding with simple primitives such as linear inequalities and the like. A global constraint is a replacement for a set of lower level constraints, where additional domain knowledge allows the application of specialized data representations and algorithms to ....

Puget, J.-F., and Leconte, M. 1995. Beyond the Glass Box: Constraints as Objects. In Proceedings of the 1995 International Logic Programming Symposium (ILPS'95), 513--527.

....of programming is being done in imperative style, it is natural to analyse in what way constraint programming can be realized within the imperative programming paradigm. Up to now, the most successful approach to imperative constraint programming is the objectoriented approach taken by ILOG Solver (Puget Leconte 1995, ILOG 1998) In this system constraints and variables are treated as objects and are defined within a C class library. Thanks to the class encapsulation mechanism and the operator overloading capability of C , the user can see constraints almost as if they were a part of the language. A ....

Puget, J.-F. & Leconte, M. (1995), Beyond the glass box: Constraints as objects, in `Proc. of the 1995 International Symposium on Logic Programming', pp. 513--527.

....performance of simple Tabu Search can be significantly increased by incorporating specific long term memory features of Guided Local Search. 2. Constraint Based Modelling of VRPs Part of the attraction of Constraint Programming is the richness of the language for specifying the model. ILOG Solver [14] allows real and integer variables to be constrained by simple bounds and linear or non linear constraints. Similarly the objective function can be complex. Very general expressions can be used as constraints. In addition to expressions involving the usual arithmetic and logical operators, complex ....

J.-F. Puget and M. Leconte. Beyond the glass-box: Constraints as objects. In Proceedings of ILPS '95. MIT Press, 1995.

....we get much better (O(N) space complexity, much smaller scheduler overhead, and the opportunity to employ a specialized, complete filtering algorithm [19] instead of merely mimicking the pairwise disequations. The need for specialized algorithms is most obvious on hard combinatorial problems [1,3,18], while the space complexity aspects can dominate on large instances of otherwise easy problems. Consequently, solvers based solely on indexicals can hardly be competitive on these classes of problems. On the other hand, indexicals admit rapid prototyping of user defined constraints: defining a ....

J.-F. Puget and M. Leconte. Beyond the glass box: Constraints as objects. In J. Lloyd, editor, Proceedings of the International Logic Programming Symposium (ILPS-95), pages 513--527, Portland, 1995.

....worsened or failed to improve the search behavior. The only useful technique was randomization, which was true for some decision points only, and the advantage decreased with the inclusion of stronger knowledge. The concept of global constraints was originally used for refinement search (e.g. [LeP94, PL95]) Transferring it to a local search context makes it possible to get an efficient handle on local search, while preserving features like declarativeness, reusability, and maintenance. Further work includes an extension to Excalibur s planning model, concepts to temporally focus the search on ....

Puget, J.-F., and Leconte, M. 1995. Beyond the Glass Box: Constraints as Objects. In Proceedings of the 1995 International Logic Programming Symposium (ILPS'95), 513--527.

....indicate that two particular loads cannot be carried on the same vehicle due to hazards in doing so, or that two remote visits must be carried out at the same time. 1 Given a language expressive enough for specifying constraints (for instance a constraint programming toolkit such as ILOG Solver [23], CHIP [34] Claire [4] etc. arbitrarily complex constraints can be specified, and it be 2 comes difficult to customize a search algorithm for the problem at hand, especially when the constraints within the problem are changing dynamically. In this paper we test general (not algorithmically ....

J.-F. Puget and M. Leconte. Beyond the glass-box: Constraints as objects. In Proceedings of ILPS '95. MIT Press, 1995.

No context found.

Puget, J-F. & Leconte, M. (1995) Beyond the glass box: constraints as objects. Proc. of International Symposium on Logic Programming (ILPS'95), MIT Press.

No context found.

Puget, J., and Leconte, M. Beyond the glass box: Constraints as objects. In Proc. International Logic Programming Symposium (1995), MIT Press, pp. 513--527.

No context found.

Puget, J.-F., Leconte, M. "Beyond the Glass Box: Constraints as Objects ". In Lloyd, J. (ed.) Proceedings of International Logic Programming Symposium (ILPS95) pp. 513-527, Portland 1995.

No context found.

Puget, J.-F., Leconte, M. "Beyond the Glass Box: Constraints as Objects". In Lloyd, J. (ed.) Proceedings of International Logic Programming Symposium (ILPS95) pp. 513-527, Portland 1995.

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