191014K02 | Day 3 Tutorial

191014K02: Day 3 Tutorial

Back to the Course Page

Definitions

Max Bisection

In the Max Bisection problem we are given a (weighted) graph $G=(V, E)$, and the objective is to find a bisection

$$V=S \cup \bar{S},|S|=|\bar{S}|=|V| / 2$$

such that the number (weight) of edges between $S$ and $\bar{S}$ is maximized.

$k$-SAT-Local Search

Given an instance of $k$-SAT, find a satisfying assignment that sets at most $d$ variables to true.

Problems

1. Come up with an instance where the majority rounding idea for the Closest String LP does not give an optimal solution. How much can you push the gap between OPT and the quality of the solution obtained by the greedy rounding.

2. Show that the majority rounding idea for the Closest String LP is a valid $2$-approximation.

3. Make the local search phase for Closest String (discussed in class) work without any knowledge of OPT (i.e, you are not allowed to guess the value of OPT).

4. Design a randomized algorithm for $k$-SAT-Local Search with running time $O(k^d)$.

5. Design a PTAS for Max Bisection on graphs of minimum degree $dn$.

6. Prove that selecting coordinates according to the normal distribution gives unifom distribution on unit sphere.

7. Prove that the projection of a random unit vector in $\mathbb{R}^d$ on any plane through the origin has a “u.a.r. direction”.