Scribd
Upload
87 views4 pages

1 Celebrity Identification Problem

This document contains a sample problem and solution for identifying a celebrity at a party using the fewest number of questions. The problem is formulated as identifying a sink vertex in a directed graph. An efficient solution is presented that uses an elimination phase to reduce the possible celebrity candidates to one, followed by a verification phase to check if that remaining candidate is indeed the celebrity. The solution makes between n and 3(n-1) queries and correctly identifies the celebrity, if one exists.

Uploaded by

jzdoog
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
87 views4 pages

1 Celebrity Identification Problem

This document contains a sample problem and solution for identifying a celebrity at a party using the fewest number of questions. The problem is formulated as identifying a sink vertex in a directed graph. An efficient solution is presented that uses an elimination phase to reduce the possible celebrity candidates to one, followed by a verification phase to check if that remaining candidate is indeed the celebrity. The solution makes between n and 3(n-1) queries and correctly identifies the celebrity, if one exists.

Uploaded by

jzdoog
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

COS 423 · Spring 2013 · Problem 0.

1 Kevin Wayne (wayne)

OUTPUT:
This document contains a sample problem and solution. Use it as a template for how
to write and format a solution in LATEX.

1 Celebrity Identification Problem


Rita, a columnist for the Daily Princetonian, is covering a party. Rita’s job is to
identify a celebrity, if one exists. A celebrity is person that is known by every other
person, but doesn’t know any of them. Rita asks questions to the guests of the
following form: Excuse me. Do you know the person over there? Assume that all of
the guests at the party are polite (even the celebrity) and answer any question with
the correct answer. Explain how Rita can identify the celebrity using as few questions
as possible.

Before looking at the solution, think about the problem on your own.

Grader’s note: there is no need to include a copy of the problem statement when you
submit your problem set. We include it here so that this document is self contained.

Collaborators: dhlarkin, sachinr Page 1 of 4


COS 423 · Spring 2013 · Problem 0.1 Kevin Wayne (wayne)

2 Mathematical Formulation
We model the relationships among the guests using a digraph G = (V, E) as follows:
There is a vertex for each of the n guests and an edge from u to v if guest u knows
guest v. We define a sink of a digraph to be a vertex with indegree n−1 and outdegree
0. A celebrity corresponds to a sink of the digraph. We note that a digraph can have
at most one sink. We also note that the edges E are given to us implicitly—we have
an oracle HasEdge(u, v) that tells us whether there is an edge from u to v and we
seek to identify a sink (if one exists) using as few calls to HasEdge(u, v) as possible.

Grader’s note: If the problem has already been stated explicitly in mathematical terms,
skip this section.

3 Brute-Force Solution
We compute the digraph G explicitly by calling HasEdge(u, v) for each potential
edge. At this point, we can check whether a vertex v is a sink by computing its
indegree and its outdegree. The digraph has at most n(n−1) edges, so the algorithms
makes Θ (n2 ) calls to HasEdge() to build the digraph. It yields little, if any, new
insight into the problem. Below, we show how to do it with at most 3(n − 1) calls to
HasEdge().

Grader’s note: Very little partial credit for a correct, but grossly, inefficient solution.
Moreover, while the solution is concise, it is missing both a formal statement and
proof of correctness and a formal statement and analysis of its efficiency.

4 An Efficient Solution
Our algorithm for finding a sink consists of two phases: in the elimination phase, we
eliminate all but one vertex s as a possible sink; in the verification phase we check
whether s is a sink.
The elimination phase maintains a list of possible sinks. Initially it contains all
n vertices. In each iteration, we remove one vertex. We exploit the following key
observation: if there is an edge from u to v, then u is not the sink; if there is not
an edge from u to v, then v is not the sink. Thus, by calling HasEdge(u, v) we can
eliminate either u or v from the list of possible sinks. We use this idea repeatedly to
eliminate all vertices but one, say s.

Page 2 of 4
COS 423 · Spring 2013 · Problem 0.1 Kevin Wayne (wayne)

Algorithm 1 Eliminating non-sinks.


procedure Eliminate(V, E)
L ← MakeList
for v ∈ V do
add(L, v)
while L contains at least two elements do
u ← Remove(L)
v ← Remove(L)
if HasEdge(u, v) then
Insert(L, v)
else
Insert(L, u)
s ← Remove(L)
return s

We now verify by brute force whether s is a sink: for every other vertex v, we
call both HasEdge(s, v) and HasEdge(v, s). If the former always returns false
and the latter always returns true, then we declare s to be the sink. Otherwise, we
conclude there is no sink.

Algorithm 2 Verifying a potential sink.


procedure Verify(V, s)
for v ∈ V \ {s} do
if HasEdge(s, v) or ¬ HasEdge(v, s) then
return false
return true

5 Correctness
Proposition 1. The eliminate-and-verify algorithm either outputs the sink (if one
exists) or reports that there is none.

Proof. During the elimination phase, we maintain the invariant that if there exists
a sink, then the sink is in the list. We omit the trivial proof by induction on the
number of iterations. Thus, when the elimination phase ends, either s is a sink or
there is no sink. This is determined by brute force during verification.

6 Analysis
Proposition 2. The eliminate-and-verify algorithm makes between n and 3(n − 1)
calls to HasEdge().

Page 3 of 4
COS 423 · Spring 2013 · Problem 0.1 Kevin Wayne (wayne)

Proof. The elimination phase requires exactly n − 1 calls to HasEdge() since each
call decreases the size of the list by one. In the verification phase, we call HasEdge()
between 1 and 2(n − 1) times, depending on whether s is a sink.

7 A Slight Improvement
We note that it is possible to save an additional blog2 nc calls in the verification
phase by not repeating any call we already asked during the elimination phase. By
maintaining the list of vertices in a FIFO queue, the sink is involved in at least blog2 nc
questions during the elimination phase.
Also, it is not hard to see that any algorithm must make at least 2(n − 1) calls
if there exists a sink because we must verify that the sink points to no vertices and
that every other vertex points to the sink.

Grader’s note: Some bonus points for a better-than-expected solution.

Grader’s note: Typically there won’t be a simple observation like this that can be
tacked onto the end, and a bonus-worthy solution will be a modification to the core of
the algorithm or analysis.

8 Appendix
A typical solution to a problem will contain a mathematical formulation and any
notation you are introducing (if needed), a description of your algorithm, a proof of
correctness, and an analysis of efficiency. Feel free to use sections and subsections to
help clarify your answer, but don’t go too crazy with them or we will not be able to
follow your argument.
It is worth noting that this sample solution contains two independent solutions,
one of which has a full analysis as well as an improvement. In most cases, you should
submit only one solution and analysis. If you submit two solutions, we will use the
worst one when assigning your score.

Page 4 of 4

You might also like