顶点着色是某种意义上的边缘着色吗?


16

我们知道,一个图的边染色G 一个特殊的图的顶点着色,即折线图L(G)G

是否有操作员图形,使得图形的顶点着色ģ 曲线图的边染色Φ ģ ?我对这样一种可以在多项式时间内构造的图算子感兴趣,即可以从G在多项式时间内获得图 Φ G ΦG Φ(G)Φ(G)G

备注:对于稳定的集合和匹配,可以询问类似的问题。中的匹配是L G )中的稳定集。是否有图运算符Ψ使得G中的稳定集与Ψ G 中的匹配?由于STABLE SET为N P -complete并且MATCHING属于P,因此假设N P不能在多项式时间内构造这样的图算子Ψ(如果存在) GL(G)ΨGΨ(G)NPPΨNPP

编辑:受@usul的答案以及@Okamoto和@King的评论的启发,我发现了一种较弱的形式:图顶点着色是定义如下的超图Φ G )的边缘着色。设定的顶点Φ ģ 是同一顶点组G ^。对于每一个顶点vģ,封闭附近Ñ ģ [ v ] = Ñ ģv { v } 。然后GG Φ(G)Φ(G)GvGNG[v]=NG(v){v}是超图的边缘Φ(G)G是超图的线图,因此顶点着色ģ被的边染色Φ ģ Φ(G)GΦ(G)

同样,对于所有答案和评论,我表示感谢,无论是否假设,我要寻找的运算符都不存在。如果我接受所有答案,那就太好了!NPP


感谢所有人的友好评论(和耐心等待!)和有用的答案。我需要时间阅读,思考,并且可能会重新焕发神采。
user13136

6
我遇到了以下由Nishizeki和Zhou于1998年提出的非常有趣的问题,该问题与您的问题以及对@TsuyoshiIto的第二条评论有某种联系: 可以将顶点着色问题“简单地”简化为边缘着色问题吗?(...)由于这两个问题都是NP完全的,由于NP完全的理论,可以通过3-SAT将其中一个问题简化为另一个问题。因此,开放的问题问...(请参阅此处
vb le

@vble:谢谢!我承认我想要“太多”。这样的运营商将解决Nishizeki和Zhou的问题。
user13136 2013年

Answers:


16

比喻 折线图,我认为您要提出以下问题:

对于每一个无向图,确实存在一个无向图G ^ ' = V 'È ',使得每个顶点v V对应于边缘v 1v 2È '和对应于边缘û Vv V份额至少一个端点当且仅当Û èG=(V,E)G=(V,E)vV(v1,v2)EuVvV(u,v)E

答案是“ 否”。考虑四个顶点树与根v有三个孩子X Ÿ ž。在G '中,我们必须具有四个边:v 1v 2x 1x 2y 1y 2z 1z 2。此外,一定是这样的,要么vGvx,y,zG(v1,v2),(x1,x2),(y1,y2),(z1,z2) v 2v1v2是每个其它三个边缘的端点(,等等)。但是,这意味着其他三个边中的至少两个必须共享一个公共端点,这违反了我们的要求,因为原始图中x y z中没有两个相邻。|{v1,v2}{x1,x2}|1x,y,z

我认为同一张图也会为您提供匹配问题的反例。


3
好点子!其实我也有同样的想法。但是也许还有另一种定义?或者我们如何才能正式证明不存在这样的算子ΦGΦ
user13136

1
@ user13136,嗯,也许有一些创造性的解决方法,但是您需要重新表述您的问题(我想我的反例是对引用框中所提问题的正式证明)。直观地讲,我认为问题在于,在沿线图方向移动时,我们需要一条边(只能连接到两个顶点)并将其变成一个顶点(可以连接到任意数量的边)-很简单。反之则相反,难度更大。
usul

2
仅添加到usul的答案中,简短的答案是“否”,因为匹配具有不一定在稳定集中存在的结构特性。例如,每个折线图也是准线和无爪的;与顶点着色相比,这确实限制了边缘着色的深度。
安德鲁·金

14

这个问题在“图形G的顶点着色是图形H的边缘着色”的含义上含混不清,但是构造一个边缘色数等于的(顶点)色数的图是NP难的。给定的图。形式上,以下关系问题是NP难的。

代表色数为边缘色数
实例:图形ģ
解决方案:图H使得边缘色数χ'(ħ)的ħ等于色数χ(G ^) of G.

这是因为 Vizing’s theorem gives a (trivial) efficient algorithm which approximates the edge chromatic number within an additive error of 1 whereas the chromatic number is hard even to approximate in various senses. For example, Khanna, Linial, and Safra [KLS00] showed that the following problem is NP-complete (and later Guruswami and Khanna [GK04] gave a much simpler proof):

3-着色与非-4-着色
实例:图形ģ
是的承诺 G是3色的。
不保证G不可四色。

这个结果足以证明我一开始就声称的NP硬度。剩下的证明可以作为练习,但这是一个提示:

锻炼身体。通过将“ 3色与非4色”还原为多项式时间函数可简化性,证明上述问题“将色数表示为边缘色数”是NP-难的。也就是说,构造两个多项式时间函数f(将图映射到图)和g(将图映射到位),使得

  • 如果G是3色图,而H是使得χ(fG))=χ'(H)的图,则gH)= 1。
  • 如果G是不可四色的图,而H是使得χ(fG))=χ'(H)的图,则gH)= 0。

参考文献

[GK04] Venkatesan Guruswami和Sanjeev Khanna。关于4色3色图的硬度。 SIAM离散数学杂志,18(1):30-40,2004 . DOI:10.1137 / S0895480100376794

[KLS00] Sanjeev Khanna, Nathan Linial, and Shmuel Safra. On the hardness of approximating the chromatic number. Combinatorica, 20(3):393–415, March 2000. DOI: 10.1007/s004930070013.


Thank you for reply! I am a little bit imprecise by formulating “vertex colorings of a graph G are edge colorings of a graph H”. What I mean is an operator Φ like the line graph operator L, but from vertex colourings to edge colourings. This is somehow more than χ(G)=χ(H).
user13136

Since VERTEX COLOURING and EDGE COLOURING are both NP-complete, we can construct, by definition, H from G in polynomial time such that χ(G)k iff χ(H)k.But such a construction need not fulfill the property for an operator Φ I am looking for. It only reduces vertex colourings to edge colourings.
user13136

1
@user13136: If a weaker requirement is impossible to satisfy, the stronger requirement is obviously also impossible. This is logic. You should understand that your planar graph example is not a counterexample to this. Deciding the 3-colorability of a given planar graph is not a weaker requirement than deciding the 4-colorability of a given planar graph; they are just different requirements. On the other hand, I already showed that what you want is impossible unless P=NP, period. But if you have trouble understanding this, I do not think there is anything I can do to help you understand.
Tsuyoshi Ito

1
If I understand the question correctly, such a map Φ doesn't exist. We don't need to refer to NP-completeness. Just consider G=K1,3 and suppose such Φ(G) exists. Since G is 2-colorable, Φ(G) should be 2-edge-colorable. This means the maximum degree of Φ(G) is at most two. Since Φ(G) has four edges, we can go through all candidates for Φ(G) (seven candidates up to isomorphism), and we will find that the family of edge colorings of Φ(G) and the family of vertex colorings of G are different. A contradiction.
Yoshio Okamoto

1
@user13136: It occurred to me that you might have been confused because I wrote only a proof idea and I left out the actual proof. I revised the answer so that it would be clear that I left out the actual proof, and added some hints for proof. If this still does not work for you, then I will give up.
Tsuyoshi Ito

9

(This is an addition to usul's answer and YoshioOkamoto's comment, rather than an answer.) It can be seen that your operation Φ exists only for those graphs G for which there is a graph G with G=L(G), i.e. G is a line graph (checkable in polytime). In this case, Φ is the "inverse line graph operator" L1, i.e. Φ(G)=G, and vertex colorings of G are edge colorings of Φ(G).

By using our site, you acknowledge that you have read and understand our Cookie Policy and Privacy Policy.
Licensed under cc by-sa 3.0 with attribution required.