## Iterated Greedy Graph Coloring and the Difficulty Landscape (1992)

Citations: | 32 - 2 self |

### BibTeX

@TECHREPORT{Culberson92iteratedgreedy,

author = {Joseph C. Culberson},

title = {Iterated Greedy Graph Coloring and the Difficulty Landscape},

institution = {},

year = {1992}

}

### Years of Citing Articles

### OpenURL

### Abstract

Many heuristic algorithms have been proposed for graph coloring. The simplest is perhaps the greedy algorithm. Many variations have been proposed for this algorithm at various levels of sophistication, but it is generally assumed that the coloring will occur in a single attempt. We note that if a new permutation of the vertices is chosen which respects the independent sets of a previous coloring, then applying the greedy algorithm will result in a new coloring in which the number of colors used does not increase, yet may decrease. We introduce several heuristics for generating new permutations that are fast when implemented and effective in reducing the coloring number. The resulting Iterated Greedy algorithm(IG) can obtain colorings in the range 100 to 103 on graphs in G 1000; 1 2 . More interestingly, it can optimally color k-colorable graphs with k up to 60 and n = 1000, exceeding results of anything in the literature for these graphs. We couple this algorithm with several other c...

### Citations

7320 |
Genetic Algorithms
- Goldberg
- 1989
(Show Context)
Citation Context ...of constraints. This approach showed no significant improvement over DSATUR using a random order. This entire project grew from an initial trivial attempt to apply the technique of Genetic Algorithms =-=[13]-=- to graph coloring. Several variations were attempted without much success, but in trying to determine what should be preserved we observed that reordering the color sets and applying the greedy algor... |

679 | Approximation algorithms for combinatorial problems - Johnson - 1974 |

577 | Where the really hard problems are
- Cheeseman, Kanefsky, et al.
- 1991
(Show Context)
Citation Context ... MAXIS found colorings ranging from 63 to 66 in the ten graph sample, and although IG reduced each coloring somewhat, no graph in the set of ten received a 60-coloring. Cheeseman, Kanefsky and Taylor =-=[7]-=- noted a similar threshold effect, stated in terms of constraint satisfaction, for three and four colorable graphs. They related this phenomenon to several different NP-complete problems. This search ... |

350 | New methods to color the vertices of a graph - Brélaz - 1979 |

229 |
Future paths for integer programming and links to artificial intelligence
- Glover
- 1986
(Show Context)
Citation Context ...ed greedy algorithm, although sometimes successful, still does not perform well on G 1000; 1 2 . Hertz and de Werra [15] describe the application of a technique called Tabu search attributed to Glover=-=[12]-=-. Using this technique they were able to color graphs in G 1000; 1 2 with 93 colors. Using a more sophisticated technique, in which they first found large independent sets using Tabu and then applied ... |

131 |
Using tabu search techniques for graph coloring
- Hertz, deWerra
- 1987
(Show Context)
Citation Context ...n some conflicts may remain. If the partition number is not the chromatic number, the results may also be an approximation to the coloring number. Example algorithms are Anti-voter[28] and Tabu search=-=[15]-=- and simulated annealing[6]. Tabu search will be described in more detail in section 6. Clique After choosing the first vertex, choose vertices with a maximal number of constraints on the colors avail... |

90 |
An upper bound for the chromatic number of a graph and its application to timetabling problems
- Welsh, Powel
- 1967
(Show Context)
Citation Context ...on the first sets may cause problems for later choices, forcing the use of unnecessary colors. Thus, these approaches produce approximatescolorings. There are many variations, heuristics and analyses =-=[29, 30, 22, 17, 16, 24, 8, 21, 20, 4, 25]-=-. Partition Partition the vertices by some means, and then attempt to remove conflicts by moving vertices from one partition to another. These methods may produce an approximately correct coloring, si... |

81 |
On colouring random graphs
- Grimmett, McDiarmid
- 1975
(Show Context)
Citation Context ... cause a conflict. For random graphs in which each vertex pair is assigned an edge with probability 1=2, it is known that asymptotically this will on average use about twice as many colors as required=-=[14]-=-. In the worst case it will use \Theta(n) times as many colors as needed[17]. Many heuristics have been proposed for ordering the vertices for coloring by the greedy algorithm. But usually the greedy ... |

66 | Almost all k-colorable graphs are easy to color
- Turner
- 1988
(Show Context)
Citation Context ...olorable graphs that even our full system cannot color well. On the other hand, most k-colorable graphs are easily colored, even for values of k well outside the ranges suggested in previous analyses =-=[27, 19]-=-. 1 Definitions Given a simple undirected graph G = (V; E), where V is a set of vertices, and E is a set of pairs of vertices called edges, a k-coloring of G is an assignment C : V ! f1 : : : kg such ... |

65 |
The chromatic number of random graphs
- Bollob'as
- 1988
(Show Context)
Citation Context ...ty p. This class of graphs has been deeply studied with respect to coloring, especially for p = 1=2. Asymptotically, for a fixed probability p, the chromatic number is known to almost surely (a.s.) be=-=[3]-=- ` 1 2 + o(1) ' log ` 1 1 \Gamma p ' n log n The greedy algorithm is known to assign approximately twice the chromatic number when coloring graphs with p = 1=2 [23, 14]. In one sense, the class of ran... |

59 | Smallest-last ordering and clustering and graph coloring algorithms - Matula, Beck - 1983 |

57 |
The complexity of near optimal graph coloring
- Garey, Johnson
- 1976
(Show Context)
Citation Context ...e a coloring assignment within less than O(n 0:4 ) [2] of the optimal coloring and are very complex. Since it is known that it is NP-hard to guarantee a coloring within a factor of two of the optimal =-=[11]-=-, this leaves a large gap in our knowledge of the difficulty of coloring and dismal prospects in the worst case. Numerous algorithms have been developed for average case coloring. The simple greedy al... |

55 |
D.: Some experiments with simulated annealing for coloring graphs
- Chams, Hertz, et al.
- 1987
(Show Context)
Citation Context ... If the partition number is not the chromatic number, the results may also be an approximation to the coloring number. Example algorithms are Anti-voter[28] and Tabu search[15] and simulated annealing=-=[6]-=-. Tabu search will be described in more detail in section 6. Clique After choosing the first vertex, choose vertices with a maximal number of constraints on the colors available to them. This is almos... |

36 |
On some properties of linear complexes
- Zykov
- 1949
(Show Context)
Citation Context ...r, we are less likely to make an error that forces bad colorings later on. We will discuss an implementation of Br'elaz's algorithm in a subsequent section. Zykov Do partial traversals of Zykov's tree=-=[31]-=-. These algorithms use the recursive structure imposed on graphs by the Zykov decompositions, but rather than do a complete search, use heuristics to attempt to find a good if not optimal coloring. We... |

32 |
Graph Coloring Algorithms
- Matula, Marble, et al.
- 1972
(Show Context)
Citation Context ...on the first sets may cause problems for later choices, forcing the use of unnecessary colors. Thus, these approaches produce approximatescolorings. There are many variations, heuristics and analyses =-=[29, 30, 22, 17, 16, 24, 8, 21, 20, 4, 25]-=-. Partition Partition the vertices by some means, and then attempt to remove conflicts by moving vertices from one partition to another. These methods may produce an approximately correct coloring, si... |

28 |
Worst-case behavior of graph coloring algorithm
- Johnson
- 1974
(Show Context)
Citation Context ...n edge with probability 1=2, it is known that asymptotically this will on average use about twice as many colors as required[14]. In the worst case it will use \Theta(n) times as many colors as needed=-=[17]-=-. Many heuristics have been proposed for ordering the vertices for coloring by the greedy algorithm. But usually the greedy algorithm is seen as a one shot approach to coloring. In this paper we will ... |

27 |
A.: Random graphs of small order
- Bollobás, Thomason
- 1985
(Show Context)
Citation Context ...ial coloring, and then proceed with our heuristics. We also explore the use of Tabu search, developing a method to combine this technique with ours by alternating between them. Bollob'as and Thomason =-=[4]-=- present an algorithm that recursively selects an approximation to the largest independent set from the uncolored vertices, and colors it. They report the best coloring results for the class G 1000; 1... |

25 |
Graph coloring using eigenvalue decomposition
- Aspvall, Gilbert
- 1984
(Show Context)
Citation Context ... good if not optimal coloring. We will not be discussing these methods further in this paper; see [10, 9] for further details. Other A few algorithms use other methods such as eigenvalue decomposition=-=[1]-=-. Most of these algorithms make a single pass over the set of vertices and edges, and produce a coloring result. Tabu and simulated annealing are notable exceptions as they may iterate a basic search ... |

15 |
An Õ(n0.4 )-approximation algorithm for 3-coloring (and improved approximation algorithms for k-coloring
- Blum
- 1989
(Show Context)
Citation Context ...etabling. The best deterministic exact algorithms are hopelessly exponential; the best known polynomial time approximation algorithms cannot guarantee a coloring assignment within less than O(n 0:4 ) =-=[2]-=- of the optimal coloring and are very complex. Since it is known that it is NP-hard to guarantee a coloring within a factor of two of the optimal [11], this leaves a large gap in our knowledge of the ... |

13 |
Probabilistic bounds and heuristic algorithms for coloring large random graphs,” Southern Methodist Univ
- Johri, Matula
- 1982
(Show Context)
Citation Context ...r IG would be superior, and whether a mix would be an improvement. Unfortunately, time considerations make this a mute question on large graphs. 9 The Maximum Independent Set Approach Johri and Matula=-=[18]-=- (see also Manvel[20]) and Bollob'as and Thomason[4] developed extremely greedy coloring algorithms. These algorithms attempt to recursively find a maximal independent set and assign a color to its ve... |

13 |
A Technique for Coloring a Graph Applicable to Large Scale Timetable Problems
- Wood
- 1969
(Show Context)
Citation Context ...on the first sets may cause problems for later choices, forcing the use of unnecessary colors. Thus, these approaches produce approximatescolorings. There are many variations, heuristics and analyses =-=[29, 30, 22, 17, 16, 24, 8, 21, 20, 4, 25]-=-. Partition Partition the vertices by some means, and then attempt to remove conflicts by moving vertices from one partition to another. These methods may produce an approximately correct coloring, si... |

8 |
Graphs with small chromatic numbers are easy to color
- Kucera
- 1989
(Show Context)
Citation Context ...olorable graphs that even our full system cannot color well. On the other hand, most k-colorable graphs are easily colored, even for values of k well outside the ranges suggested in previous analyses =-=[27, 19]-=-. 1 Definitions Given a simple undirected graph G = (V; E), where V is a set of vertices, and E is a set of pairs of vertices called edges, a k-coloring of G is an assignment C : V ! f1 : : : kg such ... |

7 |
On various algorithms for estimating the chromatic number of a graph
- Mitchem
- 1976
(Show Context)
Citation Context |

4 |
Extremely greedy coloring algorithms
- Manvel
- 1985
(Show Context)
Citation Context |

4 |
A randomised heuristical algorithm for estimating the chromatic number of a graph
- Zerovnik
- 1989
(Show Context)
Citation Context ... since on termination some conflicts may remain. If the partition number is not the chromatic number, the results may also be an approximation to the coloring number. Example algorithms are Anti-voter=-=[28]-=- and Tabu search[15] and simulated annealing[6]. Tabu search will be described in more detail in section 6. Clique After choosing the first vertex, choose vertices with a maximal number of constraints... |

3 |
A new graph coloring algorithm
- Dutton, Brigham
- 1981
(Show Context)
Citation Context ...by the Zykov decompositions, but rather than do a complete search, use heuristics to attempt to find a good if not optimal coloring. We will not be discussing these methods further in this paper; see =-=[10, 9]-=- for further details. Other A few algorithms use other methods such as eigenvalue decomposition[1]. Most of these algorithms make a single pass over the set of vertices and edges, and produce a colori... |

3 |
Worst case analysis of a graph coloring algorithm
- Spinrad, Vijayan
- 1985
(Show Context)
Citation Context ... form large cliques with respect to the vertices already chosen. Examples of these are the No Choice algorithm[27] and Br'elaz's DSATUR algorithm[5]. Counter example graphs for DSATUR are provided in =-=[26]-=-. The intuition is that if we choose vertices which are forced to be a certain color, we are less likely to make an error that forces bad colorings later on. We will discuss an implementation of Br'el... |

2 |
Coloring random graphs badly
- McDiarmid
- 1979
(Show Context)
Citation Context ...r is known to almost surely (a.s.) be[3] ` 1 2 + o(1) ' log ` 1 1 \Gamma p ' n log n The greedy algorithm is known to assign approximately twice the chromatic number when coloring graphs with p = 1=2 =-=[23, 14]-=-. In one sense, the class of random graphs is easy, since good approximations are obtained on average. Also, we do not have a specific target color to shoot for. On the other hand, they are difficult ... |

1 |
Sequential colourings of graphs
- Dunstan
- 1976
(Show Context)
Citation Context |

1 |
A Las Vegas coloring algorithm
- Ellis, Lepolesa
- 1989
(Show Context)
Citation Context ...by the Zykov decompositions, but rather than do a complete search, use heuristics to attempt to find a good if not optimal coloring. We will not be discussing these methods further in this paper; see =-=[10, 9]-=- for further details. Other A few algorithms use other methods such as eigenvalue decomposition[1]. Most of these algorithms make a single pass over the set of vertices and edges, and produce a colori... |

1 |
Some sequential graph colouring algorithms
- Sarma, Bandyopadhyay
- 1989
(Show Context)
Citation Context |