Skip to Content
Combinatorics

7. Probabilistic Method

2019-09-04Original-language archivelegacy assets may be incomplete

Existence by Probability

x,A(x)P(A(x))>0\exists x,A(x)\Leftarrow P(A(x))>0

x,A(x)P(¬A(x))<1\exists x,A(x)\Leftarrow P(\neg A(x))<1

P(A)=P(i=1nAi)>0P(i=1nAi)<1i=1nP(Ai)P(A)=P(\wedge_{i=1}^n\overline{A}_i)>0\Leftarrow P(\vee_{i=1}^nA_i)<1\Leftarrow \sum_{i=1}^nP(A_i)

Lower Bound for Ramsey's Number

  • R(k,k),R(k,k),\forall two coloring of Kn,n>R(k,k),K_n,n>R(k,k),\exists a monochromactic KkK_k
  • Erdős Theorem(1947): (nk)21Ck2<1{n\choose k}2^{1-C_k^2}<1 then it is possible to color the edges of KnK_n with two colors so that there is no monochromatic KkK_k subgraph

Paradoxical Tournament

  • orientation of the edges of the complete graph on set of vertices VV
  • kk-paradoxical: V,V=k,v\forall V, |V|=k,\exists v beats them all
  • Theorem (Erdős 1963): if (nk)(12k)nk<1{n\choose k}(1-2^{-k})^{n-k}<1, there is a tournament on nn vertices that is kk-paradoxical

The Averaging principle

x,A(x)tE(x)t\exists x,A(x)\geq t\Leftarrow E(x)\geq t

Hamilton Path in Tournament

  • Theorem (Szele 1943): There is a tournament on nn players with at least n!2(n1)n!2^{-(n-1)} Hamiltonian paths

Independent sets

  • independent set: a set of vertices with no edges between them
  • Theorem: GG has an independent set with at least n24m\frac{n^2}{4m} vertices

Coloring large-girth graphs

  • G(n,p)G(n,p): graph of nn vertices and each edge occurs at probability pp
  • girth: g(G)g(G) is the length of the shortest cycle in GG
  • χ(G)=min(CN:V[C],uvE,f(u)f(v))\chi(G)=\min(C\in\mathbb{N}|\exists :V\rightarrow [C],\forall uv\in E,f(u)\neq f(v))
  • Independent number: α(G)=max{SSV and u,vS,uv∉E}\alpha(G)=\max\{|S||S\subseteq V \text{ and }\forall u,v\in S,uv\not\in E\}
  • Lemma: χ(G)nα(G)\chi(G)\geq\frac{n}{\alpha(G)}
  • Theorem (Erdős 1959): k,,G,g(G)>,χ(G)>k\forall k,\ell,\exists G,g(G)>\ell,\chi(G)>k

Lovász Local Lemma

Consider a set of "bad" events A1,A2,,AnA_1,A_2,\cdots,A_n, suppose P(Ai)p,1inP(A_i)\leq p,1\leq i\leq n

  • Mutually independent events: P(i=1nAi)(1p)nP(\wedge_{i=1}^n\overline{A_i})\geq(1-p)^n
  • Arbirarily dependent events: P(i=1nAi)=1P(i=1n)1npP(\wedge_{i=1}^n\overline{A_i})=1-P(\vee_{i=1}^n)\geq 1-np
  • dependency graph: D(V,E)D(V,E)
    • ijE    Aiij\in E\iff A_i and AjA_j are dependent
    • d=max{deg(i)}d=\max\{\deg(i)\}
  • Lovasz Local Lemma
    • i,P(Ai)p\forall i,P(A_i)\leq p
    • ep(d+1)1ep(d+1)\leq 1 or 4dp14dp\leq 1
    • Then P(i=1nAi)>0P(\wedge_{i=1}^n\overline{A}_i)>0
  • General Lovasz Local Lemma
    • x1,,xn[0,1),i,P(Ai)xi(j,i)E(1xj)\exists x_1,\cdots,x_n\in [0,1),\forall i, P(A_i)\leq x_i\prod_{(j,i)\in E}(1-x_j)
    • Then P(i=1nAi)i=1n(1xi)P(\wedge_{i=1}^n\overline{A}_i)\geq \prod_{i=1}^n(1-x_i)