Toolbox

C1: One-Way Communication Complexity

Reminder: This post contains 1001 words · 3 min read · by Xianbin

In this post, the error of a randomized protocol is at most ϵ\epsilon where ϵ\epsilon is some constant.

One-way Set Disjointness

Theorem 1\textbf{Theorem 1}. The randomized one-way DISJ has communication complexity Ω(n)\Omega(n).

Proof. We reduce the problem DISJ to the problem INDEX. Let (X,i)(X, i) be the input instance of INDEX. For DISJ, Alice has X=XX' = X and Bob has Y=1iY' = \mathbb{1}_i where 1i=(0,0,,1,0,)\mathbb{1}_i = (0,0,\ldots,1, 0,\ldots) where only ii-th coordinate of YY' has 1. It is easy to see that DISJ (X,Y)=1X', Y') = 1 if and only if Xi=0X'_i = 0. Then we reduce every protocol for INDEX to DISJ.

One-way Gap Hamming

The gap hamming GHn\text{GH}_n is defined as follows. Alice has a string X={0,1}nX = \{0,1\}^n and Bob has a string Y={0,1}nY = \{0, 1\}^n. The goal is to determine whether XYn/2cn\lvert X- Y \vert \geq n/2 - c\sqrt{n} or XYn/2+cn\lvert X - Y \rvert \leq n/2 + c\sqrt{n} where cc is a parameter that is independent of X,Y,nX, Y, n.

Theorem 2\textbf{Theorem 2}. The randomized one-way GHn\text{GH}_n has communication complexity Ω(n)\Omega(n)

Reference

[1]. Woodruff, D.P., 2004, January. Optimal space lower bounds for all frequency moments. In SODA (Vol. 4, pp. 167-175).