B. Selman. Stoc hastic searc h and phase transitions: AI meets physic/ IJCAI, pp.998-1002, vol.1, 1995.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....our watermarking strategy is based on preprocessing of the input instance and is nonintrusive in that any existing solution method remains applicable to the augmented (watermarked) 3SAT instance. In particular, any SAT solvers from the four major classes can be used: 1) randomized local search [46]; 2) exact deterministic methods based on resolution [19] and branch and bound; 3) nonlinear programming relaxation and rounding [25] and 4) a variety of binary decision diagram based techniques [11] In our experience, many commonly encountered NP complete formulations can also be watermarked ....

B. Selman, "Stochastic search and phase transitions: AI meets physics," Proc. Int. Joint Conf. Artificial Intelligence, vol. 1, pp. 998--1002, Aug. 1995.

....strategies for the SAT problem. For example, SAT techniques have been used in testing [Ste96, Kon93] logic synthesis, and physical design [Dev89] There are at least three broad classes of solution strategies for the SAT problem. The first class of techniques are based on probabilistic search [Sil99, Sel95], the second are approximation techniques based on rounding the solution to a nonlinear program relaxation [Goe95] and the third is a great varietyofBDD based techniques [Bry95] For brevityand due to limited source code availability,we demonstrate our forensic engineering technology on the ....

B. Selman. Stochastic search and phase transitions: AI meets physics. IJCAI, pp. 998-1002, vol.1, 1995.

....on the other hand, it is likely that some relevant attributes will be omitted if the heuristic approach is taken. Our goal becomes clear, i.e. to have a reasonably fast algorithm that can find M relevant attributes with high probability. Based on the study in line with [ Cheeseman et al. 1991; Selman, 1995 ] this work proposes a probabilistic approach, in particular, a Las Vegas algorithm, that makes probabilistic choices to help guide the search more quickly to find a correct set (or sets) of M attributes. Las Vegas algorithms (LV s) use randomness to guide their search in such a way that a ....

B. Selman. Stochastic search and phase transitions: AI meets physics. In C.S. Mellish, editor, Proceedings of IJCAI95, volume 1, pages 998--1002, 1995.

....reductions extremely fast. 1000 problems maybesolved in 3 seconds or so on a notebook 486 based computer running Linux and gcc. The fast solution time seems to be due to (a) the pruning power of path consistency# (b) the now generally recognised phase transition feature of NP complete problems [Sel95] in this case, with extremely high probability, randomly generated interval networks with n:c 14 are inconsistent, whichis rapidly detected by a path consistency computation, whereas small networks are consistent with dense solution nodes, entailing that there is little backtracking on either ....

....to solve large problems, unless it can considerably reduce the time spent in path consistency calculations. Summary We knew already [LR92] that recurring path consistency computations are an effective pruning method, that practically eliminates backtracking except in the phase transition [Sel95]ofinterval constraint problems. We also knew that the phase transition for c =1laybetween 6 n 15, and that a majority of time in the algorithm was spentin path consistency computations, in particular in composition calculations. In this paper, we show that ffl the phase transition lies ....

[Article contains additional citation context not shown here]

Bart Selman. Stochastic search and phase transitions: AI meets Physics. In Proceedings of the 14th International Joint ConferenceonArtificial Intelligence (IJCAI-95),volume 1, pages 998--1002, Montr'eal, Canada, August 1995. IJCAII.

....as SAT instances. For example, SAT techniques have been used in testing [Sil97, Ste96, Cha93, Kon93] logic synthesis, and physical design [Dev89] There are at least three broad classes of solution strategies for the SAT problem. The first class of techniques are based on probabilistic search [Gu99, Sil99, Sel95, Dav60], the second are approximation techniques based on rounding the solution to a nonlinear program relaxation [Goe95] and the third is a great variety of BDD based techniques [Bry95] For brevity and due to limited source code availability, we demonstrate our forensic engineering technology on the ....

B. Selman. Stochastic search and phase transitions: AI meets physics. IJCAI, pp.998-1002, vol.1, 1995.

....typical size, say, 30 50 faculties and 100200 students each year. At the same time, combinatorial optimization has made a great progress for the past ten years. Currently a number of NP problems, such as Satisfiability, are viewed as solvable up to the size of thousands or even tens of thousands[Sel95]. From this point of view, the above size of university databases is just moderate. Nevertheless, there have been few attempts trying to fully automate combinatorial optimization over those databases, e.g. the development of optimal class schedules from university databases. The reason must be ....

B. Selman. Stochastic search and phase transitions: AI meets physics. Proc. IJCAI-95, pp.9981002, 1995.

....do not time out and engineers do it even less: after computing resources have been consumed (and paid for) some indicative result is anyway expected for action. To turn around this obstacle, approximation techniques have been developed with a particular attention to problems in NP (e.g. see [1, 8, 9]) This has led to a theoretical classification of approximability [1] and an experimental analysis of satisfiability which identifies easy and hard average cases [9] Most of these approaches [1, 9] focus on partial satisfiability (e.g. find an assignment which satisfy at least 90 of the ....

.... around this obstacle, approximation techniques have been developed with a particular attention to problems in NP (e.g. see [1, 8, 9] This has led to a theoretical classification of approximability [1] and an experimental analysis of satisfiability which identifies easy and hard average cases [9]. Most of these approaches [1, 9] focus on partial satisfiability (e.g. find an assignment which satisfy at least 90 of the clauses) and also [8] bases its analysis of entailment on the underlying satisfiability problem. An analysis of this research area reveals that, although there are plenty of ....

[Article contains additional citation context not shown here]

B. Selman. Stochastic search and phase transitions: AI meets physics. In Proc. of IJCAI-95, p. 998--1001, 1995.

.... using the extra clauses ffu 2 ; u 1 ; u 10 g;fu 14 ; u 2 ; u 8 g;fu 4 ; u 14 ; u 3 g;fu 8 ; u 12 ; u 14 gg (we pad the end of the message with an extra space to maintain three literals per clause) 4 This observation holds for the three major classes of SAT heuristics: i) random search [25, 7], ii) nonlinear programming relaxation and rounding [12] and (iii) a variety of BDD based techniques [3] 5 Artifact watermarking has been used for thousands of years. Only with the proliferation of digital media has it attracted wide research and economic interest, e.g. for protection of ....

B. Selman, "Stochastic Search and Phase Transitions: AI Meets Physics", IJCAI, 1 (1995), pp.998-1002.

....propositional logic of context [4] is NP complete [23] Although logic of context is more tractable than general ML systems (unless NP=PSPACE) it is still intractable. To turn around this general computational barrier, approximation techniques have been developed for many problems in NP (e.g. see [1, 2, 20, 26, 27]) This has led to a theoretical classification of approximability [2] and an experimental analysis of satisfiability which identifies easy and hard average cases [27] Most of these approaches [2, 27] focus on partial satisfiability (e.g. find an assignment for at least 90 of the clauses) and ....

.... this general computational barrier, approximation techniques have been developed for many problems in NP (e.g. see [1, 2, 20, 26, 27] This has led to a theoretical classification of approximability [2] and an experimental analysis of satisfiability which identifies easy and hard average cases [27]. Most of these approaches [2, 27] focus on partial satisfiability (e.g. find an assignment for at least 90 of the clauses) and [20, 26] base their analysis of entailment on the underlying satisfiability problem. In the quest for intelligent and effective applications a question is important: ....

[Article contains additional citation context not shown here]

B. Selman. Stochastic search and phasetransitions: AI meets physics. In Proc. of IJCAI-95, pages 998--1001, 1995.

....probabilistic analysis [ 7 ] In particular, it is an open question in the DL community how the inference algorithms behave on average. The recent insights that have been gained into the understanding of the SAT problem were due to the performance analysis of random instances of the problem [ 9 ] Average case analysis requires precise models of the distribution of input instances. Ideally, those models should reflect real world distributions. However, since we usually do not have a good understanding of realworld distributions, average case analysis has to rely on random samples of ....

....easy when the problem instances have a regular structure whose basic building blocks are not interrelated. It is the case for the SAT problem: the instances of SAT are clauses made of literals which are either propositional variables or their negations. In the so called fixed clause length model [ 9 ] each clause is generated by randomly selecting k variables among n, each of which is negated with probability 0:5. This test model has been adapted to the setting of propositional modal logic and its equivalent DL ALC ( 4; 5 ] and also [ 1 ] in order to capture hard instances for ....

B. Selman. Stochastic search and phase transitions: AI meets physics. In Proceedings of AAAI-94, 1994.

No context found.

B. Selman, Stochastic search and phase transitions: AI meets physics, in: Proc. of the 14th International Joint Conference on Artificial Intelligence (IJCAI-95), Montreal, Canada (1995).

No context found.

B. Selman. Stoc hastic searc h and phase transitions: AI meets physic/ IJCAI, pp.998-1002, vol.1, 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