Schuurmans, D., & Southey, F. (2001). Local search characteristics of incomplete SAT procedures. Artificial Intelligence, 132(2), 121--150.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....is based on a tabu search procedure and uses a similar, but slightly more complicated penalty update scheme. Additionally, DLM periodically and deterministically invokes a smoothing mechanism that decreases all clause penalties by a constant amount. The Smoothed Descent and Flood (SDF) approach [11] introduced a more complex smoothing method, and the concept of multiplicative penalty updates, which evolved into the Exponentiated Sub Gradient (ESG) method [12] Our Scaling and Probabilistic Smoothing (SAPS) 7] method improved upon the ESG approach; SAPS will be described in detail in Section ....

D. Schuurmans and F. Southey. Local search characteristics of incomplete SAT procedures. In Proc. AAAI-2000, pp. 297--302, AAAI Press, 2000.

....[15] is based on a tabu search procedure and uses a similar, but slightly more complicated weight update scheme. Additionally, DLM periodically and deterministically envokes a smoothing mechanism that decreases all clause weights by a constant amount. The Smoothed Descent and Flood (SDF) approach [10] introduced a more complex smoothing method, and the concept of multiplicative weight updates. The most recent and best performing DLS algorithm for SAT is the Exponentiated Sub Gradient (ESG) method [11] ESG, described in more detail in the next section, reaches or exceeds the performance of the ....

D. Schuurmans, and F. Southey. Local search characteristics of incomplete SAT procedures. In Proc. AAAI-2000, pp. 297-302, AAAI Press, 2000.

....local search studies, namely, GSAT [15] and WalkSAT [12] 14] They both have many variations, for example, GWSAT [14] GSAT Tabu [11] of GSAT, and Novelty, R Novelty [12] Novelty , and R Novelty [6] of WalkSAT. Recently, two more e ective local search methods have been proposed, they are: SDF [13] and UnitWalk [5] Generally speaking, SDF makes a greedy descent on its objective function similar to other local search methods. But a special feature of SDF is that it not only counts the number of unsatis ed clauses, but also considers how strongly a clause is satis ed. In addition, SDF ....

D. Schuurmans and F. Southey, Local search characteristics of incomplete SAT procedures, Arti cial Intelligence, vol. 132, no. 2, pp. 121-150, 2001.

....been proposed to solve SATs. The conventional methods can be categorized into two types. The rst type is based on DPLL procedure [3, 4] for example, GRASP [13] REL SAT [11] SATZ [12] and TABLEAU [2] etc. Another type is based on local search [5, 20] such as GSAT [20] WalkSAT [14, 19] SDF [17, 18] and UnitWalk [8, 9] All these conventional methods are sequential. Recently, two distributed multi agent based approaches have been proposed to solve SATs, i.e. MASSAT [10] and MultiDB [7] In both distributed approaches, the authors employed a multi agent system to solve SATs in which agents ....

D. Schuurmans and F. Southey. Local search characteristics of incomplete sat procedures. Arti cial Intelligence, 132(2):121-150, 2001.

....been proposed to solve SATs. The conventional methods can be categorized into two types. The rst type is based on DPLL procedure [3, 4] for example, GRASP [13] REL SAT [11] SATZ [12] and TABLEAU [2] etc. Another type is based on local search [5, 20] such as GSAT [20] WalkSAT [14, 19] SDF [17, 18] and UnitWalk [8, 9] All these conventional methods are sequential. Recently, two distributed multi agent based approaches have been proposed to solve SATs, i.e. MASSAT [10] and MultiDB [7] In both distributed approaches, the authors employed a multi agent system to solve SATs in which agents ....

D. Schuurmans and F. Southey. Local search characteristics of incomplete sat procedures. In Proceedings of the Seventeenth National Conference on Arti cial Intelligence (AAAI-2000.

....Gent and Walsh [18] gives a overview of some complete and incomplete algorithms for propositional satis ability search. We feel that our work is focused more on local search methods, and that we go more into detail on some of the more interesting techniques known today. Schuurmans and Southey [48] analyzes many of the local search procedures described in this report, and de nes characteristic properties for the algorithms. Also, they introduce a new local search algorithm for SAT based on their results. The work we present here give a more thorough look on the di erent techniques used in ....

....section we will take a closer look on the most in uential techniques. 2.3.1 Flat moves One obvious strategy to escape benches is simply to allow moves that neither increases nor decreases the cost of the search. This approach is used by Wu and Wah [65] Shang and Wah [53] Schuurmans and Southey [48], Hoos and Stutzle [25] Galinier and Hao [13] Walser [61] Frank et al. 12] Cha and Iwama [3] Morris [38] Selman et al. 52] and Ekelin and Olovsson [9] The rst algorithms for SAT local search simply gave up when discovering that no assignment in the neighborhood had a lower cost than the ....

[Article contains additional citation context not shown here]

Schuurmans, D., and Southey, F. Local search characteristics of incomplete SAT procedures. In AAAI/IAAI (2000), pp. 297-302.

....modi es it step by step until it nds a satisfying assignment. If the random walk is long enough, it is restarted from another initial assignment. There is a number of experimentally studied SAT local search algorithms, for example, GSAT [34] GWSAT (aka GSAT w) 33] HSAT [6] HWSAT [7] SDF [32], IDB [29] WalkSAT [33] and various algorithms within WalkSAT framework such as WalkSAT TABU, Novelty, R Novelty [23] Novelty and R Novelty [14] Some of these algorithms need restarts because there are initial assignments such that the probability that the random walk (if allowed to run in ....

Schuurmans, D. and F. Southey: 2000, `Local search characteristics of incomplete SAT procedures'. In: Proc. of AAAI'2000. pp. 297-302.

.... For the past decade, various types of stochastic local search (SLS) methods have been applied very successfully to the propositional satisfiability problem (SAT) These include the GSAT and WalkSAT families of algorithms [19, 4, 14] as well as several other algorithms based on similar ideas [7, 20, 21, 16, 17]. GSAT and WalkSAT algorithms have been extensively studied in the literature, and include some of the best performing SAT algorithms known to date [11, 17] Compared to other state of the art SAT algorithms (e.g. Satz [13] these methods are rather simplistic and it is not well understood how ....

D. Schuurmans and F. Southy. Local search characteristics of incomplete sat procedures. Artificial Intelligence, 132:121--150, 2000.

....it step by step until it nds a satisfying assignment. If the random walk is long enough, it is restarted from another initial assignment. There is a number of experimentally studied SAT local search algorithms, for example, GSAT [SLM92] GWSAT (aka GSAT w) SKC94] HSAT [GW93] HWSAT [GW95] SDF [SS00], WalkSAT [SKC94] and various algorithms within WalkSAT framework such as WalkSAT TABU, Novelty, R Novelty [MSK97] Novelty and R Novelty [Hoo99] Some of these algorithms need restarts because there are initial assignments such that the probability that the random walk (if allowed to run in ....

D. Schuurmans and F. Southey. Local search characteristics of incomplete SAT procedures. In Proc. of AAAI'2000, pages 297-302, 2000.

....compared to other contemporary incomplete SAT solvers, it dominates on satis able instances of aim [AIM96] ssa [Lar92] par [JT96] and ii [KKRR92] series. It is also able to solve some of the qg [ZH94] instances, all satis able Velev s microprocessor veri cation instances from series SSS.1.0 [Velb] SSS.1.0a [Velc] and SSS SAT.1.0 [Vela] and most instances from VLIW SAT.1.1 [Veld] Our algorithm is inspired by two randomized algorithms having the best current worst case upper bounds ( PPSZ98, PPZ97] and [Sch99, SSW01] In our paper we combine the main ideas of these ....

....contemporary incomplete SAT solvers, it dominates on satis able instances of aim [AIM96] ssa [Lar92] par [JT96] and ii [KKRR92] series. It is also able to solve some of the qg [ZH94] instances, all satis able Velev s microprocessor veri cation instances from series SSS.1. 0 [Velb] SSS.1.0a [Velc] and SSS SAT.1.0 [Vela] and most instances from VLIW SAT.1.1 [Veld] Our algorithm is inspired by two randomized algorithms having the best current worst case upper bounds ( PPSZ98, PPZ97] and [Sch99, SSW01] In our paper we combine the main ideas of these algorithms in one ....

[Article contains additional citation context not shown here]

D. Schuurmans and F. Southey. Local search characteristics of incomplete SAT procedures. In Proc. of AAAI'

....by UnitWalk on various benchmarks from [5] and its running time. We also present the number of substitutions made, which characterizes the current implementation better. For the comparision, we present the data for several other algorithms: GSAT [15] WalkSAT [14] Novelty, R Novelty [8] SDF [13], IDB (aka CLS) 10] ....

D. Schuurmans and F. Southey. Local search characteristics of incomplete SAT procedures. In Proc. of AAAI'2000, pages 297-302.

....of Walksat is the exact form and parameterization of the search function. Our earlier work [62] showed that the optimal mix of greedy to non greedy steps for many functions is one in which the ratio of the mean to the variance of the number of unsatisfied clauses is minimized. Schuurman [82] describes qualities of depth , mobility , and coverage that a good search function should maximize, and uses these qualities to design a new search function that outperforms the simple probabilistic mix model. Hoos [38] provides evidence that any purely deterministic search function will fail ....

D. Schuurmans and F. Southey. Local search characteristics of incomplete SAT procedures. In Proceedings of the Seventeenth National Conference on Articial Intelligence (AAAI-

....guide the search can easily be incorporated into the search process. The approach that was used here is based on [5] Similar techniques were applied in [7, 9, 11] The use of weights for a re structuring of the neighborhood has been identified as a powerful strategy. Work in this area includes [2, 8, 10]. We do not as is these approaches only change the weights when a local optimum is reached because the conventional definition of a local optimum is not applicable. Instead, the weights are adapted when a respective decision point is visited again. This re weighting procedure does not only ....

Schuurmans, D., and Southey, F. 2000. Local search characteristics of incomplete SAT procedures. AAAI-2000, 297--302.

.... (which we describe below) Our algorithm arises from the observation that the most recent strategies developed for the SAT problem have begun to use an analog of subgradient optimization as their core search strategy; in particular the DLM system of [Wu and Wah, 2000] and the SDF system of [Schuurmans and Southey, 2000] . Interestingly, these are among the most effective methods for finding satisfying assignments for CNF formulae, and yet they appear to be recapitulating a forty year old idea in OR going back to [Everett, 1963] One of our contributions is to show that, indeed, a straightforward subgradient ....

....ver, there are subtle differences between recent SAT procedures and the basic subgradient optimization approach outlined above. The DLM system of [Wu and Wah, 2000] explicitly uses a Lagrangian, but the multiplier updates follow a complex system of ad hoc calculations. 2 The SDF system of [Schuurmans and Southey, 2000] is simpler (albeit slower) but includes several details of uncertain significance. However, the basic simplicity of this approach offers many useful hypotheses for our current investigation. Given the clear connection between recent SAT procedures and the traditional subgradient optimization ....

[Article contains additional citation context not shown here]

D. Schuurmans and F. Southey. Local search characteristics of incomplete SAT procedures. In Proceedings AAAI-00, pages 297--302, 2000.

.... (which we describe below) Our algorithm arises from the observation that the most recent strategies developed for the SAT problem have begun to use an analog of subgradient optimization as their core search strategy; in particular the DLM system of [Wu and Wah, 2000] and the SDF system of [Schuurmans and Southey, 2000] (see also [Thornton and Sattar, 1999; Frank, 1997; Davenport et al. 1994] Interestingly, these are among the most effective methods for finding satisfying assignments for CNF formulae, and yet they appear to be recapitulating a forty year old idea in OR going back to [Everett, 1963] Below ....

....However, there are subtle differences between recent SAT procedures and the basic subgradient optimization approach outlined above. The DLM system of [Wu and Wah, 2000] explicitly uses a Lagrangian, but the multiplier updates follow a complex system of ad hoc calculations. 2 The SDF system of [Schuurmans and Southey, 2000] is simpler (albeit slower) but includes several details of uncertain significance. Nevertheless, the simplicity of this latter approach offers useful hypotheses for our current investigation. Given the clear connection between recent SAT procedures and the traditional subgradient optimization ....

[Article contains additional citation context not shown here]

D. Schuurmans and F. Southey. Local search characteristics of incomplete SAT procedures. In Proc. AAAI-00, pages 297--302, 2000.

.... (which we describe below) Our algorithm arises from the observation that the most recent strategies developed for the SAT problem have begun to use an analog of subgradient optimization as their core search strategy; in particular the DLM system of [Wu and Wah, 2000] and the SDF system of [Schuurmans and Southey, 2000] . Interestingly, these are among the most effective methods for finding satisfying assignments for CNF formulae, and yet they appear to be recapitulating a thirty year old idea in OR going back to [Everett, 1963] One of our contributions is to show that, indeed, a straightforward subgradient ....

....However, there are subtle differences between recent SAT procedures and the basic subgradient optimization approach outlined above. The DLM system of [Wu and Wah, 2000] explicitly uses a Lagrangian, but the multiplier updates follow a complex system of ad hoc calculations. 2 The SDF system of [Schuurmans and Southey, 2000] is simpler (albeit slower) but includes several details of uncertain significance. However, the basic simplicity of this latter approach offers many interesting hypotheses for our investigation. Given the clear connection between recent SAT procedures and the traditional subgradient ....

[Article contains additional citation context not shown here]

D. Schuurmans and F. Southey. Local search characteristics of incomplete SAT procedures. In Proceedings AAAI-00, pages 297--302, 2000.

....We show that SDF exhibits uniformly good depth, mobility, and coverage values, and consequently achieves good search performance (in terms of the number of search steps required) on a large collection of benchmark satis ability problems. This paper revises and extends the earlier work reported in [20]. 2 Local search procedures In this paper we investigate several dominant local search procedures from the current literature. Although many of these strategies appear to be only super cial variants of one another, they demonstrate dramatically di erent problem solving performance and (as we ....

D. Schuurmans and F. Southey. Local search characteristics of incomplete SAT procedures. In Proceedings of the Seventeenth National Conference on Articial Intelligence (AAAI-2000), pages 297-302, 2000.

No context found.

Schuurmans, D., & Southey, F. (2001). Local search characteristics of incomplete SAT procedures. Artificial Intelligence, 132(2), 121--150.

No context found.

Schuurmans, D., Southey, F.: Local search characteristics of incomplete SAT procedures. In: Proceedings of the 1tth National Conference on Artificial Intelligence (AAAI'00), Austin, Texas, AAAI Press/The MIT Press (2000) 297--302

No context found.

D. Schuurmans and F. Southey. Local search characteristics of incomplete SAT procedures. In Proceedings of the Seventeenth National Conference on Artificial Intelligence (AAAI-00), pages 297--302, 2000.

No context found.

Dale Schuurmans and Finnegan Southey. Local search characteristics of incomplete SAT procedures. In Proceedings of the Seventeenth National Conference in Artificial Intelligence (AAAI'00), pages 297--302, 2000.

No context found.

D. Schuurmans, and F. Southey. Local search characteristics of incomplete SAT procedures. In Proc. AAAI-2000, pp. 297--302, AAAI Press, 2000.

No context found.

Dale Schuurmans and Finnegan Southey. Local search characteristics of incomplete SAT procedures. In Proceedings of the Seventeenth National Conference in Artificial Intelligence (AAAI'00), pages 297--302, 2000.

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