#### DMCA

## Selecting the Right Heuristic Algorithm: Runtime Performance Predictors (1996)

Venue: | Advances in Artificial The Eleventh Biennial Conference of the Canadian Society for Computational Studies of Intelligence |

Citations: | 19 - 2 self |

### Citations

730 | A new method for solving hard satisfiability problems,
- Selman, Levesque, et al.
- 1992
(Show Context)
Citation Context ... small differences in an algorithm, or the representation of a problem, can make a big difference in performance. As an example, we describe our recent experience comparing min-conflicts [9] and GSAT =-=[17]-=-, two iterative-repair algorithms. Min-conflicts begins with an initial assignment. On each iteration, min-conflicts randomly chooses a variable that is in conflict and assigns it the value that minim... |

683 | Where the Really Hard Problems Are;
- Cheeseman, Kanefsky, et al.
- 1991
(Show Context)
Citation Context ...agation and iterative repair tabu search. 2 Finding the Right Heuristic Algorithm is Important Recently, researchers have been interested in identifying "hard" constraint satisfaction proble=-=ms (e.g., [1, 10]). Informally, an problem is hard if no -=-algorithm can solve "most" of its instances "quickly", for suitable definitions of "most" and "quickly". Similarly, a CSP problem is trivial if most CSP algorit... |

679 | Tabu search—Part I
- Glover
- 1989
(Show Context)
Citation Context ...olor will give a solution.) 2. GSAT requires a tabu list. (Selman used a tabu list of length 20. A tabu list of length k keeps track of the last k moves, so they are not repeated by the repair process=-=[2, 3]-=-. This can help the algorithm cope with local minima and plateaus.) 3. GSAT's variable-ordering strategy helps significantly. Armed with this knowledge, we were able to create a variation of min-confl... |

449 | Algorithms for Constraint-Satisfaction Problems: A Survey
- Kumar
- 1992
(Show Context)
Citation Context ...ms for our demonstration: constraintpropagation methods in a backtracking framework and tabu search in an iterative repair framework. 4.1 Estimated Total Constraint Checks Constraint-propagation (CP) =-=[12, 7]-=- is frequently used in conjunction with backtracking search. The idea is simple: whenever a variable is instantiated with a new value, we can prune the domains (the possible values) of the uninstantia... |

387 | Tabu search — part II
- Glover
- 1990
(Show Context)
Citation Context ...olor will give a solution.) 2. GSAT requires a tabu list. (Selman used a tabu list of length 20. A tabu list of length k keeps track of the last k moves, so they are not repeated by the repair process=-=[2, 3]-=-. This can help the algorithm cope with local minima and plateaus.) 3. GSAT's variable-ordering strategy helps significantly. Armed with this knowledge, we were able to create a variation of min-confl... |

312 |
and easy distributions of SAT problems
- Hard
- 1992
(Show Context)
Citation Context ...agation and iterative repair tabu search. 2 Finding the Right Heuristic Algorithm is Important Recently, researchers have been interested in identifying "hard" constraint satisfaction proble=-=ms (e.g., [1, 10]). Informally, an problem is hard if no -=-algorithm can solve "most" of its instances "quickly", for suitable definitions of "most" and "quickly". Similarly, a CSP problem is trivial if most CSP algorit... |

228 | Contradicting conventional wisdom in constraint satisfaction.
- Sabin, Freuder
- 1994
(Show Context)
Citation Context ...this sense, each successive CP heuristic is more powerful than the preceding one. However, more powerful does not necessarily result in better overall performance. In fact, recently Sabin and Freuder =-=[14]-=- have shown that the relative performance of these techniques is problem-dependent. There is a simple model that allows us to understand the basic effects of CP. Simply put, CP techniques increase eff... |

183 |
HSTS: Integrating Planning and Scheduling
- Muscettola
- 1994
(Show Context)
Citation Context ...e [5] found a backtrack version that works well as well. While N-queens is not intrinsically interesting, it is notable because it has long served as a benchmark for AI search methods. -- Researchers =-=[11, 4, 16]-=- have recently identified heuristic techniques that quickly solve a job shop problem originally proposed by Sadeh [15]. For instance, Johnston found that a combination of a dispatch heuristic and iter... |

134 | Look-Ahead Techniques for MicroOpportunistic Job Shop Scheduling.
- Sadeh
- 1991
(Show Context)
Citation Context ...se it has long served as a benchmark for AI search methods. -- Researchers [11, 4, 16] have recently identified heuristic techniques that quickly solve a job shop problem originally proposed by Sadeh =-=[15]-=-. For instance, Johnston found that a combination of a dispatch heuristic and iterative repair search solved most of the instances almost immediately. In each case, benchmark problems that were meant ... |

131 |
Estimating the efficiency of backtrack programs,
- Knuth
- 1975
(Show Context)
Citation Context ...tal number of constraint checks that would be required if the entire tree were to be searched, which we refer to as EstTCC. The procedure is based on an iterative sampling approach described by Knuth =-=[6]-=- for estimating the size of the search tree, which we adapt for use with depth-first search, as described below. Let CC(n) be the number of constraint checks that occur at node n, and k(n) be the bran... |

62 | Micro-Opportunistic Scheduling: The Micro-Boss Factory Scheduler,
- Sadeh
- 1994
(Show Context)
Citation Context ...e [5] found a backtrack version that works well as well. While N-queens is not intrinsically interesting, it is notable because it has long served as a benchmark for AI search methods. -- Researchers =-=[11, 4, 16]-=- have recently identified heuristic techniques that quickly solve a job shop problem originally proposed by Sadeh [15]. For instance, Johnston found that a combination of a dispatch heuristic and iter... |

43 |
Tree Search and Arc Consistency in Constraint Satisfaction Algorithms
- Nadel
- 1988
(Show Context)
Citation Context ...ms for our demonstration: constraintpropagation methods in a backtracking framework and tabu search in an iterative repair framework. 4.1 Estimated Total Constraint Checks Constraint-propagation (CP) =-=[12, 7]-=- is frequently used in conjunction with backtracking search. The idea is simple: whenever a variable is instantiated with a new value, we can prune the domains (the possible values) of the uninstantia... |

31 | Analysing a heuristic strategy for constraint satisfaction scheduling
- Johnston, Minton
- 1994
(Show Context)
Citation Context ...e [5] found a backtrack version that works well as well. While N-queens is not intrinsically interesting, it is notable because it has long served as a benchmark for AI search methods. -- Researchers =-=[11, 4, 16]-=- have recently identified heuristic techniques that quickly solve a job shop problem originally proposed by Sadeh [15]. For instance, Johnston found that a combination of a dispatch heuristic and iter... |

20 | Integrating heuristics for constraint satisfaction problems: A case study.
- Minton
- 1993
(Show Context)
Citation Context ...nder-appreciated. The difficulty is aggravated when an algorithm incorporates a variety of heuristic mechanisms, as is often the case, since these can interact in unexpected ways. For example, Minton =-=[8]-=- found that different backtracking algorithms were appropriate for two distributions of a CSP called Minimal Maximum Matching. The two backtracking algorithms used different variable and value orderin... |

18 |
Transformational approach to scheduling
- Smith
- 1992
(Show Context)
Citation Context ...ed quickly, but only if the right CSP method is applied. Some recent studies provide examples of challenging problems --- problems that were considered hard until the right method was found: -- Smith =-=[18]-=- describes a DARPA scheduling problem for which he synthesized an algorithm that was orders of magnitude faster than previous methods. By making heavy use of constraint propagation techniques, Smith's... |

14 |
An almost perfect heuristic for the n nonattacking queens problem
- Kale
- 1990
(Show Context)
Citation Context ...gorithm completely eliminated search in some cases. -- Minton et al. [9] have demonstrated that an iterative repair method solves instances of the N-queens problem easily, even if n = 10 6 , and Kale =-=[5]-=- found a backtrack version that works well as well. While N-queens is not intrinsically interesting, it is notable because it has long served as a benchmark for AI search methods. -- Researchers [11, ... |

8 |
Superlinear speedup in state-space search
- Rao, Kumar
- 1988
(Show Context)
Citation Context ...ng much better than others on that instance. One possible solution is to run a variety of algorithms in parallel. This can lead to a significant improvement in performance, e.g., super-linear speedup =-=[13]-=-, but typically resource limitations prevent us from running more than a few algorithms in parallel. Here we consider an approach, runtime performance prediction (RPP ), whereby we monitor an algorith... |

1 |
Minimizing conficts: a heuristic repair method for constraint satisfaction and scheduling problems
- Minton, Johnston, et al.
- 1992
(Show Context)
Citation Context ...thm that was orders of magnitude faster than previous methods. By making heavy use of constraint propagation techniques, Smith's algorithm completely eliminated search in some cases. -- Minton et al. =-=[9]-=- have demonstrated that an iterative repair method solves instances of the N-queens problem easily, even if n = 10 6 , and Kale [5] found a backtrack version that works well as well. While N-queens is... |