Minton, S. (1996). Automatically configuring constraint satisfaction programs: A case study. Constraints, 1 (1), 7--43.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....heuristic uses a normal representation. This allows easy comparisons as well as efficient combinations. A combination of heuristics acts as a parameter addressed to a search scheme or an exploration strategy. The link between the model and the heuristic component allows heuristic instantiation [Minton 96] 2.3 Search schemes A search scheme defines a structured view of a search space. We can define both local search and tree search spaces. Tree search is implemented by defining an ordered set of choice points. Each choice point refers to a sub set of variables presented in the CP model. Local ....

S. Minton. Automatically Configuring Constraint Satisfaction Programs: A Case Study. Constraints, 1(1), 1996.

....and static nature of the solution. Artificial intelligence planners like the General Problem Solver [Newell and Simon, 1963] Prodigy [Veloso et al. 1995] and SIPE [Wilkins, 1984] are described as domain independent planner systems. The problem is that generality is at odds with efficiency [Minton, 1996]. These systems provide a general, domain independent, lower level reasoning mechanism, that solve prob lems by manipulating a rich representation that encodes all the application domain knowledge needed to solve the problem. These systems are based on the assumption of separating problem ....

S. Minton. Automatically configuring constraint satisfaction programs: A case study. Constraints: An International Journal, pages 7 43, January 1996.

....al. 1993] formalized a general problem of selecting among alternative problemsolving methods and used dynamic programming to solve some special cases. Minton developed an inductive learning system that configures constraint satisfaction programs, by selecting among alternative search strategies [ Minton, 1996; Allen and Minton, 1996 ] Hansen and Zilberstein [1996] studied trade offs between running time and solution quality in simple any time algorithms, and designed a dynamic programming technique for deciding when to terminate the search. Mouaddib and Zilberstein [1995] developed a similar ....

....general problem of selecting among alternative problemsolving methods and used dynamic programming to solve some special cases. Minton developed an inductive learning system that configures constraint satisfaction programs, by selecting among alternative search strategies [ Minton, 1996; Allen and Minton, 1996 ] Hansen and Zilberstein [1996] studied trade offs between running time and solution quality in simple any time algorithms, and designed a dynamic programming technique for deciding when to terminate the search. Mouaddib and Zilberstein [1995] developed a similar technique for hierarchical ....

Steven Minton. Automatically configuring constraint satisfaction programs: A case study. Constraints: An International Journal, 1:7--43, 1996.

....on hard random 3 SAT instances from the phase transition region and SAT encoded problems from other domains can be characterised by exponential distributions. Generalising the individual instance behaviour on a whole problem class is also a main motivation for approaches like the MULTITAC system [23] which automatically configures search algorithms for a given problem class based on a training set of typical instances. From the results of our empirical investigation we also derived a number of both theoretically and practically interesting consequences. Moreover, our results give 14 These ....

S. Minton. Automatically Configuring Constraint Satisfaction Programs: A Case Study. Constraints, 1(1), 1996.

....of complexity to the modelling task. For instance, the choice of the constraint formulation is strongly affected by the choice of the representation of variables and values and of the solution methods. In addition, the performance of our solution methods is sensitive to the problem instances [13, 11]. Thus, modelling combinatorial optimisation problems so as to solve them in more efficient ways is a major challenge for constraint programming (CP) A research direction on redundant modelling is emerging within the CP community, especially for some classes of problems where a dual viewpoint ....

S. Minton. Automatically configuring constraint satisfaction programs: A case study. Constraints 1(1--2):7--43, 1996.

....optimisation can be achieved before solving time (that is, when the actual instance of the problem is still unknown) and that alternate models need to be considered. Table 1 compares various existing CSP solver model transformation approaches, namely Smith s KIDS [Smi:90] Minton s MULTI TAC [Min:96], Ellman s DA MSA [EKBA:98] and the reformulation framework proposed by Borrett and Tsang [BT:xx] The comparison is in terms of five features: 6 JFPLC 2001. KIDS MULTI TAC DA MSA B T us Time compile compile compile solve compile Instances not explicit distribution distribution current one ....

....is rather the task of a programming environment, taking a set of training instances as further input, and our advocated reformulation tool proposes just that. Indeed, notice that our other rules propose classical reformulations. Related to our work are Smith s KIDS [Smi:90] Minton s MULTI TAC [Min:96], Ellman s DA MSA [EKBA:98] and the reformulation framework proposed by Borrett and Tsang [BT:xx] A comparison between these approaches and ours has been presented in Section 3. As future work, we will implement a prototype reformulation tool, working under the scenario shown in Section 3. We ....

S. Minton. Automatically configuring constraint satisfaction programs: A case study. Constraints 1(1--2):7--43, 1996.

....to the following. As shown by Tsang et al. 17] there is no universally best solver for all instances of all problems. Thus, we are only told that a particular solver is best for the particular instances used by researchers to carry out their experiments. Therefore, as also noticed by Minton [14], the performance of solvers is instance dependent, i.e. for a given problem a solver can perform well for some (distributions on the) instances, but very poorly on others. In such a case, conventional wisdom suggests joining the competitors, although we propose a novel way of interpreting this ....

....of solver components to application domains [17] However, we here focused on just one application domain (or: class of problems) as well as on just the effect of VVO heuristics while keeping the solver otherwise constant. Also closely related to our work is Minton s multi tac system [14], which automatically synthesises an instance distribution specific solver, given a highlevel model of some CSP and a set of training instances (or a generator thereof) His motivation also was that heuristics depend on the distribution of instances. However, we differ from his approach in various ....

S. Minton. Automatically configuring constraint satisfaction programs: A case study. Constraints 1(1--2):7--43, 1996.

....a particular instance and the distribution of on a class of instances are two different distributions. Algorithm portfolio design (Huberman, Lukose, Hogg 1997; Gomes Selman 1997) aims at combining several algorithms by running them in parallel or by interleaving them on a single processor. (Minton 1996) addresses the problem of specializing general constraint satisfaction algorithms and heuristics for a particular application. Conclusion and future work We have proposed a simple method for selecting a Branch and Bound algorithm from among a set of promising ones. It is based on the estimation ....

Minton, S. 1996. Automatically Configuring Constraint Satisfaction Programs : A Case Study. Constraints 1:7--43.

.... of selecting among alternative problem solving methods and used dynamic programming to solve some special cases of this problem (Russell et al. 1993) Minton developed an inductive learning system that configures constraintsatisfaction programs, by selecting among alternative search strategies (Minton 1996). Hansen and Zilberstein studied trade offs between running time and solutionquality in simple any time algorithms, and designed a dynamic programming technique for deciding when to terminate the search (Hansen and Zilberstein 1996) Mouaddib and Zilberstein developed a similar technique for ....

Minton, S. 1996. Automatically configuring constraint satisfaction programs: A case study. Constraints: An International Journal 1:7--43.

....possibilities for future work arise. There are likely to be many issues arising when trying to implement improvements to the subsearch, in particular in trying to implement the exploitation of sub CSP structure as discussed in Chapter IX. Maybe, we could also use a CSP engine such as Multi tac [47] that could take the many calls to the subsearch engines as learning examples, and so produce specialized methods to solve the sub CSPs. However, here we just discuss a few of the issues arising. 11.1 Implications for Problem Encodings In SATPLAN, a lot of effort has gone into manipulating the ....

Steven Minton. Automatically configuring constraint satisfaction programs: A case study. Constraints, 1(1):7--43, 1996.

....due to two reasons. First, as mentioned by Tsang et al. 10] there is no universally best heuristic for all problems. Thus, we are only able to learn that a particular heuristic is best for the particular benchmarks used by researchers to carry out their experiments. Second, as noticed by Minton [8], the performance of heuristics is instancedependent, i.e. for a given problem a heuristic can perform well for some distributions on the instances, but very poorly on other distributions. To understand our terminology, note that the phrase problem class here refers to a whole set of related ....

.... CSP models are initially written in a very expressive, purely declarative, typed, first order set constraint logic programming language, such as our proposal in [3] here called esra, which is being designed to be higher level than even opl [13] Using program synthesis techniques such as those in [11, 8, 2], we can automatically compile esra programs into lower level languages such as clp(FD) or opl. The purpose of this paper is not to discuss how this can be done, nor the syntax and semantics of esra. In the class of subset decision problems, a subset S of a given finite set T has to be found, ....

[Article contains additional citation context not shown here]

S. Minton. Automatically configuring constraint satisfaction programs: A case study. Constraints 1(1--2):7--43, 1996.

....heuristic for repairing constraint violations. Minton et al. 13] empirically demonstrated that min conflicts is considerably more e#cient than traditional constructive backtracking methods. In order to improve the search performance researchers have proposed many di#erent heuristic algorithms [11]. 325 326 META HEURISTICS 98: THEORY APPLICATIONS The stochastic methods are proposed, in general, to solve large scale contraint satisfaction problems. Using an appropriate heuristic algorithm can make a tremendous di#erence in the time required to solve a CSP, 11] Evolutionary methods are ....

....heuristic algorithms [11] 325 326 META HEURISTICS 98: THEORY APPLICATIONS The stochastic methods are proposed, in general, to solve large scale contraint satisfaction problems. Using an appropriate heuristic algorithm can make a tremendous di#erence in the time required to solve a CSP, [11]. Evolutionary methods are based on the evolution theory and they are in the category of stochastic search methods for optimization problems [10] It has been applied to solving Constraint Satisfaction Optimization Problems [20] and CSP [2, 3, 14, 15, 16] These approaches have concentrated on ....

Minton S. Automatically configuring constraint satisfaction programs: A case study. In Constraints, 1996.

.... see Freuder et al. 1995) Closely 22 connected to our work is the study of search control and various constraint processing techniques (Dechter 1991, Freuder et al. 1995, Ginsberg and Geddis 1991, Kondrak and van Beek 1995) and the work on selecting appropriate search heuristics (see e.g. Minton 1996). Figure 9 shows the performance of our various strategies on the problem of finding quasigroups with no values pre assigned. From the four strategies, we see that the ones that have a stochastic element clearly outperform the completely deterministic procedure ( SS ) This confirms our initial ....

Minton, S. (1996) Automatically configuring constraint satisfaction programs: A case study. Constraints, 1 (1).

No context found.

Minton, S. 1996. Automatically configuring constraint satisfaction programs: A case study. Constraints 1(1).

....for efficiency instead of optimality, can yield a domain independent solution for initial plan generation. We also plan to develop a system that can automatically learn the optimal planner configuration for a given domain and problem distribution in a manner analogous to Minton s Multi TAC system (Minton 1996). Our system would perform a search in the configuration space of the PbR planner proposing candidate sets of rewriting rules and different search methods. By testing each proposed configuration against a training set of simple problems, the system would hillclimb in the configuration space in ....

Minton, S. 1996. Automatically configuring constraint satisfaction programs: A case study. Constraints 1(1).

No context found.

Minton, S. (1996). Automatically configuring constraint satisfaction programs: A case study. Constraints, 1 (1), 7--43.

No context found.

Steven Minton. Automatically configuring constraint satisfaction programs: A case study. Constraints: An International Journal, 1(1--2):7--43, 1996.

No context found.

Steven Minton. Automatically configuring constraint satisfaction programs: A case study. Constraints, 1(1):7--43, 1996.

No context found.

Minton, S.: Automatically configuring constraint satisfaction programs: A case study. Constraints 1 (1996) 7--43

No context found.

Minton, S.: Automatically configuring constraint satisfaction programs: A case study. Constraints 1 (1996) 7--43

No context found.

S. Minton. Automatically configuring constraint satisfaction programs: A case study. Constraints, 1(1/2):7--43, 1996. 228

No context found.

S. Minton, `Automatically Configuring Constraint Satisfaction Programs: A Case Study', Constraints, 1(1/2), 7--43, (1996).

No context found.

Minton, S.: Automatically Configuring Constraint Satisfaction Programs: A Case Study. Constraints 1 (1996) 7-43

No context found.

Minton, S., Automatically Configuring Constraint Satisfaction Programs: A Case Study. Constraints, 1996. 1(1).

No context found.

S. Minton. Automatically configuring constraint satisfaction programs: A case study. Constraints 1(1--2):7-- 43, 1996.

First 50 documents

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