Scribd
Upload
20 views

Npcomplete

This document outlines the steps to prove that a problem X is NP-complete. It involves showing that X is in NP and that X is NP-hard by reducing a known NP-complete problem Y to X. The document provides an example proof that the problem of Independent Set (IS) is NP-complete by first showing IS is in NP, and then reducing 3-SAT to IS through a polynomial-time reduction that transforms 3-SAT instances into IS instances while preserving solvability.

Uploaded by

Victor Manuel Cervantes Salido
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online on Scribd
20 views

Npcomplete

This document outlines the steps to prove that a problem X is NP-complete. It involves showing that X is in NP and that X is NP-hard by reducing a known NP-complete problem Y to X. The document provides an example proof that the problem of Independent Set (IS) is NP-complete by first showing IS is in NP, and then reducing 3-SAT to IS through a polynomial-time reduction that transforms 3-SAT instances into IS instances while preserving solvability.

Uploaded by

Victor Manuel Cervantes Salido
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

CS 482 Summer 2005

Proving a Problem is NP-Complete

To prove a problem X is NP-complete, you need to show that it is both in NP and


that it is NP-Hard. Steps 2 through 5 seek to accomplish the latter.

Step 1: Show that X is in NP. We want to argue that there is a polytime verifier
for X. In other words, for any yes instance of X, there exists a certificate that
the verifier will accept, and for any no instance of X, there is no certificate that
the verifier will accept. Both the size of the certificate and the running time of
the verifier must be polynomial. Often this step is very brief, but necessary.

Step 2: Pick a known NP-complete problem. State what problem Y you are re-
ducing to X. You need to show that Y ≤p X. You may use any problem Y
which we have proved in class to be NP-complete, as well as any problem you
have proved to be NP-complete on the homework assignments. Some problems
will be far easier to use than others in your proof.

Step 3: Construct an algorithm to solve Y given an algorithm to solve X.


You need to show that any instance of Y can be solved using a polynomial number
of operations, and a polynomial number of calls to a black box that can solve X.
It is very easy to get mixed up and instead prove that X ≤p Y . Unfortunately,
this is not what you want to show (we already know that Y is NP-complete).

Step 4: Prove the correctness of your algorithm. This has 2 parts: You want
to show that given a yes instance of Y your algorithm returns “yes”. Furthermore,
you want to show that given a no instance of Y , your algorithm returns “no”.
It is always trivial to come up with an algorithm that satisfies just one of these
two conditions. We want something that satisfies both.

Step 5: Polytime and wrap-up. Finally, you need to conclude that since your algo-
rithm runs in polynomial time, Y ≤p X. Since Y is NP-complete, X is NP-Hard,
and since we also have shown that X is in NP, X is in fact NP-complete.

An Example: INDEPENDENT SET (IS) is NP-complete


First, IS is in NP, since given any set S we can check in polytime that S is independent
and that |S| = k. So for a yes instance, we simply use an independent set of size k.

1
And for a no instance, it is clear that no such set exists. Therefore IS is in NP. Now
we will show that IS is NP-Hard via reduction to 3-SAT.

Suppose we have an instance F = C1 ∧ C2 ∧ . . . ∧ Cm , where Ci is the disjunction


of 3 variables, drawn from x1 , x2 , . . . , xn and their negations, x1 , x2 , . . . , xn . We create
the graph G as follows:

For each variable in each clause, create a node, which we will label with the name
of the variable. Therefore there may be multiple nodes with the label xi or xi , if these
variables appear in multiple clauses.For each clause, add an edge between the three
nodes corresponding the variables from that clause. We will call these three nodes and
three edges a “clause gadget”. Finally, for all i, add an edge between every pair of
nodes with one labeled xi and the other labeled xi . Now use our black box for IS to
determine whether or not there is an independent set of size m in G. If there is, return
yes. Otherwise, return no.

We must now show that this algorithm correctly determines whether or not F has
a satisfying assignment. Suppose F has a satisfying assignment A. Then at least one
variable in each clause is satisfied by A. Define S to be a set of nodes in G found by
selecting one of the satisfied variable nodes in each clause gadget. Since we picked one
node for each clause, there are clearly m nodes in S. Furthermore, since we only select
a single node for each gadget, independence is not violated within any clause gadget.
Finally, independence is not violated between clauses, since the only edges between
clauses go between nodes with labels xi and xi , and A can not have satisfied both of
these, so we never would have selected both. Therefore S is an independent set of size
k, so our algorithm will return yes.

Now suppose that our algorithm returns yes. We now need to show that there is a
satisfying assignment A. Since our algorithm returned yes, there must be an indepen-
dent set S of size m on G. Since S is independent, at most one node in each clause
gadget must be used by S. But in fact, since there are exactly m clause gadgets, S
must contain exactly one node from each clause gadget. Since S is independent, no
pair of nodes xi and xi are ever both selected for S. Consider the following assignment.
Set A(xi ) = T if xi ∈ S and set A(xi ) = F otherwise. Observe that this is a truth
assignment, since all variables are assigned either T or F , but not both. Furthermore,
A satisfies F , since by our construction, A satisfies each clause.

Therefore we have shown that 3-SAT ≤p IS. Thus IS is NP-hard, and since we have
shown IS to be in NP, IS is NP-complete.

You might also like