Computational problem

In theoretical computer science, a computational problem is a mathematical object representing a collection of questions that computers might want to solve. For example, the problem of factoring

"Given a positive integer n, find a nontrivial prime factor of n."

is a computational problem. Computational problems are one of the main objects of study in theoretical computer science. The field of algorithms studies methods of solving computational problems efficiently. The complementary field of computational complexity attempts to explain why certain computational problems are intractable for computers.

A computational problem can be viewed as an infinite collection of instances together with a solution for every instance. For example in the factoring problem, the instances are the integers n, and solutions are prime numbers p that describe nontrivial prime factors of n.

It is conventional to represent both instances and solutions by binary strings, namely elements of {0, 1}*. For example, numbers can be represented as binary strings using the binary encoding. (For readability, we identify numbers with their binary encodings in the examples below.)

Types of computational problems

A decision problem is a computational problem where the answer for every instance is either yes or no. An example of a decision problem is primality testing:

"Given a positive integer n, determine if n is prime."

A decision problem is typically represented as the set of all instances for which the answer is yes. For example, primality testing can be represented as the infinite set

L = {2, 3, 5, 7, 11, ...}

In a search problem, the answers can be arbitrary strings. For example, factoring is a search problem where the instances are (string representations of) positive integers and the solutions are (string representations of) collections of primes.

A search problem is represented as a relation over consisting of all the instance-solution pairs, called a search relation. For example, primality can be represented as the relation

R = {(4, 2), (6, 2), (6, 3), (8, 2), (8, 4), (9, 3), ...}

which consist of all pairs of numbers (n, p), where p is a nontrivial prime factor of n.

A counting problem asks for the number of solutions to a given search problem. For example, the counting problem associated with primality is

"Given a positive integer n, count the number of nontrivial prime factors of n."

A counting problem can be represented by a function f from {0, 1}* to the nonnegative integers. For a search relation R, the counting problem associated to R is the function

fR(x) = |{y: (x, y) ∈ R}|.

An optimization problem asks for finding the "best possible" solution among the set of all possible solutions to a search problem. One example is the maximum independent set problem:

"Given a graph G, find an independent set of G of maximum size."

Optimization problems can be represented by their search relations.

Promise problems

In computational complexity theory, it is usually implicitly assumed that any string in {0, 1}* represents an instance of the computational problem in question. However, sometimes not all strings {0, 1}* represent valid instances, and one specifies a proper subset of {0, 1}* as the set of "valid instances". Computational problems of this type are called promise problems.

The following is an example of a (decision) promise problem:

"Given a graph G, determine if every independent set in G has size at most 5, or G has an independent set of size at least 10."

Here, the valid instances are those graphs whose maximum independent set size is either at most 5 or at least 10.

Decision promise problems are usually represented as pairs of disjoint subsets (Lyes, Lno) of {0, 1}*. The valid instances are those in LyesLno. Lyes and Lno represent the instances whose answer is yes and no, respectively.

Promise problems play an important role in several areas of computational complexity, including hardness of approximation, property testing, and interactive proof systems.

References