I just started reading Vazirani's book "Approximation Algorithms". It is legally available online here.

On page 5 (23 in the pdf), it says that the following decision problems are in NP:

  • Is the size of the minimum vertex cover in $G$ at most $k$?
  • Is the size of the maximum matching in $G$ at least $l$?

I can't see why this is true. If, say, the first problem is in NP, then each valid instance (graph) has a Yes certificate which proves that the graph has a minimum vertex cover of size at most $k$.

But, how can one deduce from a certificate something regarding the size of the minimum vertex cover?

  • $\begingroup$ This book is still on my todo list, but I have to comment you probably will want an (lower level) book on optimization as reference as it seems to assume you're comfortable with non-approximation algorithms already. You might want to take a look at this freely available pdf by Alexander Schrijver: homepages.cwi.nl/~lex/files/dict.pdf Chapter 6 covers some of the main concepts in the introduction of Vazirani a more accesable way I think. $\endgroup$ – Thomas Bosman Aug 17 '15 at 21:49

Any vertex cover of size $k$ is a witness that the smallest one has at most size $k$. Hence, the questions

Is there a vertex cover of size at most $k$?


Does a smallest vertex have size at most $k$?

are equivalent.

| cite | improve this answer | |
  • 1
    $\begingroup$ Additional exercise, related to the question: Given an algorithm which decides if a vertex cover of a given size exists, how many times would you need to run this algorithm to find the size of the minimal cover? $\endgroup$ – Pseudonym Aug 16 '15 at 21:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.