M. Yokoo. Weak-Commitment Search for Solving Constraint Satisfaction Problems. In Proceedings of the 12th National Conference on Artificial Intelligence (AAAI '94); Vol. 1, pages 313--318, Seattle, WA, USA, July 31 - August 4 1994.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....Figure 11: Semantic Branching example (e) 4. No good Recording No good recording (also called no good learning) is a powerful pruning technique for solving general CSPs [Dechter 1990; Frost and Dechter 1994; Ginsberg and McAllester 1994; Schiex and Veffaillie 1994; Schiex and Veffaillie 1994; Yokoo 1994; Dechter and Frost 1999] and SAT problems [Roberto J. Bayardo and Schrag 1977] In this section, we adapt this technique to DTP solving. Intuitively, a no good is an assignment of the variables that cannot lead to a solution, and is thus either an induced or explicit constraint of the CSP. It is ....

Yokoo, M. (1994). Weak-commitment search for solving constraint satisfaction problems. National Conference in Artificial Intelligence (AAAI-94). 49

....in two search spaces: a solution space and a prioritisation space. Both searches in uence each other: each solution is analysed and used to change the prioritisation, which guides the search strategy used to nd the next solution, found by restarting the search. Yokoo s Weak Commitment Search [39] (WCS) greedily constructs consistent partial assignments. On reaching a dead end it randomly restarts, and it uses learning to maintain completeness. Richards Richards [32] describe a SAT algorithm called learn SAT based on WCS. None of these algorithms performs random backtracks on a small ....

M. Yokoo. Weak-Commitment Search for Solving Constraint Satisfaction Problems. AAAI Press 1994, pp. 313-318.

....icts. More precisely, the ith ranked value where i = int(log 2 (1=r) w) r is a random number uniformly distributed in [0; 1] and w is some xed weighting. 2.4. 5 Weak commitment The algorithm weak commitment (WC) is an extension of the informed backtracking algorithm introduced by Yokoo in [66]. In the weak commitment 37 algorithm, all variables are given tentative values as in the informed backtracking algorithm, and variables are added one by one to the consistent partial solution. Where a variable choice is never abandoned in informed backtracking (unless it turns out to be ....

Yokoo, M. Weak-commitment search for solving constraint satisfaction problems. In AAAI, Vol. 1 (1994), pp. 313-318. 61

....the CB algorithms (other than CB MC, discussed below) and IDB FC MD performs like LS MC. The new backtracking (BH) and value ordering (VH) heuristics further boost performance, making IDB the best reported algorithm in terms of backtracks; it also beats another hybrid called Weak Commitment Search [49] which requires approximately 35 steps for large n [34] However, in terms of CPU time IDB scales more poorly than CB FC. The time per backtrack for both scales roughly linearly with n, but we found that IDB FC MD takes approximately 3.6ns per backtrack, while CB FC MD takes 0.16ns (measured by ....

....backtracking algorithm that uses conflict analysis to unassign variables leading to constraint violation, and a heuristic similar to VH that restores assignments after backtracking. Iterative Sampling [28] restarts a constructive search every time a dead end is reached. Weak Commitment Search [49] builds consistent partial assignments, using the min conflict heuristic to guide value selection. On reaching a dead end it restarts and uses learning to avoid redundant search. Learn SAT [40] is based on Weak Commitment Search. Bounded Backtrack Search [21] is a hybrid of Iterative Sampling and ....

M. Yokoo, Weak-commitment search for solving constraint satisfaction problems, in: Proceedings of the Twelfth National Conference on Artificial Intelligence (AAAI Press, 1994) pp. 313--318.

....The two phase algorithm of Zhang Zhang [28] searches a space of partial variable assignments, alternating backtracking search with stochastic local search on the same data structure. It can be tuned to di erent problems by spending more time in either phase. Yokoo s Weak Commitment Search [26] (WCS) greedily constructs consistent partial assignments. On reaching a dead end it randomly restarts, and uses learning to maintain completeness. Richards Richards [23] describe a SAT algorithm called learn SAT based on WCS. Shaw [25] describes a vehicle routing algorithm called Large ....

M. Yokoo, Weak-commitment search for solving constraint satisfaction problems, in: Proceedings of the Twelfth National Conference on Arti- cial Intelligence, AAAI Press, 1994, pp. 313-318. 18

....optimization problems they can reach a far better quality in a given time frame. But, local search algorithms cannot guarantee that they find a solution, and may be unable to find one. And thus, they are not the panacea. Several works have studied cooperation between local and systematic search [2,13,17, 36,43,44,48,54].Those hybrid approaches have led to good results on large scale problems. Three categories of hybrid approaches can be found in the literature: performing a local search before or after a systematic search; performing a systematic search improved with a local search at some point of the ....

M. Yokoo, Weak-commitment search for solving constraint satisfaction problems, in: Proc. AAAI-94, Seattle, WA, 1994, pp. 313--318.

....IDB algorithm scales much better than all the DFS algorithms, and IDB3 performs like MCHC. The additional backtracking and value ordering heuristics further boost performance, making IDB the best reported algorithm in terms of backtracks; it also beats another hybrid called Weak Commitment Search [Yokoo, 1994] which requires approximately 35 steps for large n [Pothos and Richards, 1995] In terms of execution time IDB is also efficient, each backtrack taking a n = 10 n = 100 n = 1000 alg steps succ steps succ steps succ DFS1 81.0 100 9929 1 DFS2 25.4 100 7128 39 98097 3 DFS3 14.7 100 1268 ....

....Backtracking that increases the flexibility in choice of backtracking variable. However, they note that to achieve total flexibility while preserving completeness would require exponential memory, and they recommend a less flexible version using only polynomial memory. Weak Commitment Search [Yokoo, 1994] builds consistent partial assignments, using greedy search (the minconflict heuristic) to guide value selection. On reaching a dead end it restarts and uses learning to avoid redundant search. Local Changes [Verfaillie and Schiex, 1994] is a complete backtracking algorithm that uses conflict ....

M. Yokoo. Weak-Commitment Search for Solving Constraint Satisfaction Problems. Proceedings of the Twelfth National Conference on Artificial Intelligence, AAAI Press 1994, pp. 313--318.

....perspective on the results arising from the study of structured insoluble problems. Here we compare learn SAT and relsat. 1.5 Empirical Results 1.5. 1 Preliminary Experiments The performance of RestartRepair LbM enhanced with minimising conflicts heuristic is shown to outperform weak commitment [Yokoo, 1994], which was previously shown to outperform weight on similar problems. In contrast to the simple dead end learning of weak commitment, which is found to degrade performance dramatically in many cases, learning by merging in RestartRepair LbM never does so significantly. More importantly, as ....

....i.e. when the partial labelling cannot be consistently extended to a selected variable, the search restarts with the empty labelling. The notion of restarting each time a dead end is discovered has been used in different ways, e.g. in iterative sampling [Langley, 1992] and in weak commitment [Yokoo, 1994]. These are discussed in Section 2.6. Typically, repair procedures work on complete labellings, where every variable is assigned a value. Variables whose assignments are involved in violated constraints are repaired or re assigned in an attempt to fix these inconsistencies. In the context of ....

[Article contains additional citation context not shown here]

M. Yokoo, Weak-commitment search for solving constraint satisfaction problems, Proceedings AAAI-94, 1994

....is made without using information gathered in a previous run of the algorithm. An example for this approach is the restart from a randomly generated solution. In the other case information from a previous run is used as e.g. in the clause weighting scheme used in [12] or weak commitment search [20]. In exploration within a run it is essential to allow worsening neighborhood moves. Here basically the procedures differ in the criteria by which worsening moves are made or accepted. e.g. in random walk with a certain probability some random move is made. In Simulated Annealing a proposed ....

M. Yokoo. Weak-commitment Search for Solving Constraint Satisfaction Problems. In Proceedings of AAAI'94, pages 313--318, 1994.

....have in recent years become a growing research area. A great variety of algorithms exist and we have investigated some common algorithms. Their similarities as well as differences have been highlighted. Local search methods such as GCSP[5] min conflict hill climbing[11] and weak commitment[25], have been implemented. We have evaluated their performance on particular configuration problems which is a class of real world problems that with advantage can be expressed as constraint satisfaction problems. In this the configuration problem domain has itself been examined regarding general ....

....some constraint For each value v in the domain of x Do Change x to v Compute nr unsat xv as the number of constraints that are unsatisfied The value v which minimizes nr unsat xv is chosen for x Pseudo code 2.6: Min conflict variable value selection Weak commitment search. Weak commitment search [25] is a hybrid between local search and constructive search. Unlike the previous algorithms, weak commitment is complete. The completeness is achieved by the use of memory to record nogoods, i.e. assignments that don t lead to a solution. An assignment is also separated into two parts: vars done, ....

M. Yokoo. Weak-commitment search for solving constraint satisfaction problems. Proceedings AAAI'94, 1994.

.... Laguna 1993) They perform a probabilistic exploration of the search space and therefore cannot guarantee to find a solution, but may be far more efficient (wrt reponse time) than systematic ones to find a solution. Several works have studied cooperation between local and systematic search (Yokoo 1994; Pesant Gendreau 1996; David 1997; Schaerf 1997; Gervet 1998; Richards Richards 1998) Those hybrid approaches have led to good results on large scale problems. Three categories of hybrid approaches can be found in the literature: 1. performing a local search before or after a systematic ....

Yokoo, M. 1994. Weak-commitment search for solving constraint satisfaction problems. In Proceedings of AAAI.

....to reduce the number of constraint violations. One of the algorithms presented was a backtrack algorithm that repaired regions of constraint violation in the complete assignment. Newer methods also employed the notion of partially consistent, tentative assignments to uninstantiated variables [33, 34]. These methods also started from a partially consistent assignment and reduced constraint violations until global feasibility was obtained. 30] used probabilistic analysis to show that a random probe assignment obtained before search improves 25 20 50 100 500 1000 7 5 3 1 0 10 ....

Makoto Yokoo. Weak-commitment search for solving constraint satisfaction problems. In AAAI-94, pages 313-318, Seattle, WA, August 1994.

....computes a maximal solution of the problem of variables Vars using Goal2, uses Goal1 for the proof of optimality. sets max sol( Vars) equivalent to sets max sol(Vars; sets labeling(Vars) sets labeling(Vars) Using the semiring Sets, we have implemented a version of the Weak Commitment Search [9]. 25 wcs( Vars, Values) assigns a value for each finite domain variable in list Vars using list Values (both lists should have the same length) as starting values. This predicate should not be inserted in a constraint block. Finally, note that: this predicate is not backtrackable, this ....

Makoto Yokoo. Weak-Commitment Search for Solving Constraint Satisfaction Problems. In Proceedings of AAAI'94, Seattle, USA.

....solution. The interest of local algorithms (e.g. Tabu search [12] GSAT [25] is that, following local gradients in the search space, they may be far more efficient (wrt reponse time) than systematic ones to find a solution. Several works have studied cooperation between local and systematic search [6,8,20,22,23,30]. Those hybrid approaches have led to good results on large scale problems. Three categories of hybrid approaches can be found in the literature: Preprint submitted to Elsevier Preprint 17 September (1) performing a local search before or after a systematic search; 2) performing a systematic ....

M. Yokoo. Weak-commitment search for solving constraint satisfaction problems. In Proceedings of AAAI, 1994.

....escape from local minima [11, 21] Another well known stochastic algorithm is GSAT [18] which was designed to apply to Boolean satisfaction problems but has been successfully applied to several other CSPs. Both systematic and stochastic algorithms may be augmented by various forms of learning [1, 6, 7, 8, 15, 19, 20]: accumulating knowledge during search to avoid revisiting points in the search space. Some stochastic algorithms accumulate sufficient knowledge to ensure completeness, in which case they may be thought of as unusually flexible systematic algorithms. A currently active area of research is the ....

....case they may be thought of as unusually flexible systematic algorithms. A currently active area of research is the creation of hybrid algorithms, with the aim of combining the scalability and flexibility of non systematic algorithms with the constraint handling abilities of systematic algorithms [7, 8, 13, 15, 17, 20]. This paper is a further contribution to this field. 3 RAD backtracking In this section we describe our cut down version of Dynamic Backtracking, and how we extend it to an algorithm for solving COPs. Dynamic Backtracking and its hybrids are described in terms of nogoods, which play no part in ....

[Article contains additional citation context not shown here]

M. Yokoo. Weak-Commitment Search for Solving Constraint Satisfaction Problems. Proceedings of the 12th National Conference on Artificial Intelligence 1, AAAI Press, 1994, pp. 313--318.

....[6] Schiex Verfaillie [17, 18] Ginsberg [8] and Frost Dechter [7] These investigate no good learning in the context of backtrack search. In contrast, we apply no good learning in the context of non systematic search. The origins of learn SAT lie in ng backmarking [14] and weak commitment [19]. Ng backmarking is based on a very weak form of search driven mainly through learning by merging. Weak commitment uses a stronger form of search with weaker no good learning. Learn SAT combines the search of weak commitment with learning by merging. Before describing the components of learn SAT, ....

M. Yokoo, Weak-commitment search for solving constraint satisfaction problems. Proceedings AAAI-94,

.... RgRmNt TLL: 2 RgRmNt TLL: 1 RgRmNt TLL: 0 0 100 200 300 400 500 600 700 800 0 5 10 15 20 25 30 Inconsistency Time (sec) RgRmRt TLL: 10 RgRmRt TLL: 5 RgRmRt TLL: 4 RgRmRt TLL: 3 RgRmRt TLL: 2 RgRmRt TLL: 1 RgRmRt TLL: 0 Figure 9: Results for Tabu List Lengths of 0, 1, 2, 3, 4, 5, and 10 search [Yok94], coalition forming [HT95] and the well known min conflicts heuristic [MJPL92] for binary CSPs, with its extension and generalization by genet [DTWZ94] The most important difference of our work to all these approaches is the ability of global constraints to include constraint specific search ....

Yokoo, M. 1994. Weak-commitment Search for Solving Constraint Satisfaction Problems. In Proceedings of the Twelfth National Conference on Artificial Intelligence (AAAI-94), 313-- 318.

.... new violations) 14] iv) weak commitment search, which starts with an initial full assignment and then tries to extend a partial set of variables that is completely consistent, and which restarts this process whenever a variable is found that cannot be added without engendering an inconsistency [15]. In the original version, discarded partial solutions were added to the problem as nogoods; however, it was found here that the proliferation of nogoods slowed down processing considerably even with an efficient storage and lookup mechanism. Hence, this feature is not used in the present version, ....

M. Yokoo. Weak-commitment search for solving constraint satisfaction problems. In Proceedings AAAI-94, pages 313--318, 1994. This article was processed using the L a T E X macro package with LLNCS style

....but their effects remain unclear. Alternative attempts to enhance the performance of systematic search procedures have lead to a number of hybrid algorithms, e.g. arbitrary choice dynamic backtracking, partial order dynamic backtracking [ Ginsberg and McAllester, 1993 ] and weak commitment search [ Yokoo, 1994 ] Initial experiments with these algorithms have yielded successful results where pure local search procedures have failed. Another line of inquiry addresses the distribution of problem instances with respect to search effort. Research in this area has revealed that CSPs exhibit an ....

.... VarsLeft (X,d) X v, v in Values and v minimises conflicts PartialSol PartialSol [ X,v) WCS (VarsLeft,PartialSol,Nogoods) END procedure Figure 2. The weak commitment search algorithm. WCS is a hybrid of constructive and local search (Figure 2) It can be summarised briefly as follows [ Yokoo, 1994 ] Given: An initial assignment for the problem variables Procedure: Commit to a partial solution as long it can be extended consistently (weak commitment to the current partial solution) When a dead end is reached, record the partial solution as a no good and start constructing a new partial ....

Yokoo M. Weak-commitment Search for Solving Constraint Satisfaction Problems. In Proceedings of AAAI,1994.

No context found.

Makato Yokoo. Weak-commitment search for solving constraint satisfaction problems. In Proceedings of the Weak-Commitment Search for Solving Constraint Satisfaction Problems, pages 313-318, 1994.

....while the completeness of the algorithm is guaranteed. Furthermore, we describe how this asynchronous backtracking algorithm can be modified into a more efficient algorithm called asynchronous weak commitment search, which is inspired by the weakcommitment search algorithm for solving CSPs [9]. The main characteristic of this algorithm is as follows. ffl Agents can revise a bad decision without an exhaustive search by changing the priority order of agents dynamically. In the asynchronous backtracking algorithm, the priority order of agents is determined, and each agent tries to find ....

....that are not compatible with the agent view. Therefore, each agent x i needs to record at most jD i j nogoods, where jD i j is the number of possible values of x i . V. Asynchronous Weak Commitment Search In this section, we briefly describe the weak commitment search algorithm for solving CSPs [9], and describe how the asynchronous weak commitment search algorithm is constructed by modifying the asynchronous backtracking algorithm. 9 We should mention that the waytodetermine that agents as a whole have reached a stable state is not contained in this algorithm. To detect the stable state, ....

[Article contains additional citation context not shown here]

M. Yokoo, "Weak-commitment search for solving constraint satisfaction problems", in Proceedings of the Twelfth National ConferenceonArtificial Intelligence, 1994, pp. 313--318.

....search, that is, the same variable can be revised again and again. Therefore, these algorithms can be efficient, but their completeness cannot be guaranteed. There exist several hybrid type algorithms of backtracking and iterative improvement. For example, the weak commitment search algorithm [24] is based on the min conflict backtracking. However, in this algorithm, when for one variable no value satisfies all of the constraints with the partial solution, instead of changing one variable value, the whole partial solution is abandoned. The search process is restarted using the currentvalue ....

Yokoo, M.: 1994, `Weak-commitment Search for Solving Constraint Satisfaction Problems'. In: Proceedings of the Twelfth National Conference on Artificial Intelligence. pp. 313--318.

....while the completeness of the algorithm is guaranteed. Furthermore, we describe how this asynchronous backtracking algorithm can be modified into a more efficient algorithm called asynchronous weak commitment search, which is inspired by the weak commitment search algorithm for solving CSPs [9]. The main characteristic of this algorithm is as follows: # Agents can revise a bad decision without an exhaustive search by changing the priority order of agents dynamically. In the asynchronous backtracking algorithm, the priority order of agents is determined, and each agent tries to find a ....

....that are not compatible with the agent view. Therefore, each agent x i needs to record at most D i nogoods, where D i is the number of possible values of x i . 5ASYNCHRONOUS WEAK COMMITMENT SEARCH In this section, we briefly describe the weak commitment search algorithm for solving CSPs [9] and describe how the asynchronous weak commitment search algorithm is constructed by modifying the asynchronous backtracking algorithm. 5.1 Weak Commitment Search Algorithm In the weak commitment search algorithm, all the variables have tentative initial values. A consistent partial solution is ....

[Article contains additional citation context not shown here]

# M. Yokoo, "Weak-Commitment Search for Solving Constraint Satisfaction Problems," Proc. 12th Nat'l Conf. Artificial Intelligence, pp. 313--318, 1994.

No context found.

M. Yokoo. Weak-Commitment Search for Solving Constraint Satisfaction Problems. In Proceedings of the 12th National Conference on Artificial Intelligence (AAAI '94); Vol. 1, pages 313--318, Seattle, WA, USA, July 31 - August 4 1994.

No context found.

Makoto Yokoo. Weak-Commitment Search for Solving Constraint Satisfaction Problems. In Proceedings of the 12th National Conference on Artificial Intelligence (AAAI-94); Vol. 1, pages 313--318, Seattle, WA, USA, July 31 - August 4 1994. AAAI Press, 1994.

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