Computability theory

Computability theory is that part of the theory of computation dealing with which problems are solvable by algorithms (equivalently, by Turing machines), with various restrictions and extensions. Computability theory addresses four main questions:
 What problems can Turing machines solve?
 What other systems are equivalent to Turing machines?
 What problems require more powerful machines?
 What problems can be solved by less powerful machines?
See the article on theory of computation for a chart showing which classes of problems are subsets of other classes.
Contents 
What problems can Turing machines solve?
Not all problems can be solved. An undecidable problem is one that cannot be solved by any algorithm, even given unbounded time and memory. Many undecidable problems are known. For example, the Entscheidungsproblem (German for "decision problem") is this: given a statement in firstorder predicate calculus, decide whether it is universally valid. Church and Turing independently proved this is undecidable. The halting problem is: given a program and inputs for it, decide whether it will run forever or will eventually halt. Turing proved this is also undecidable. A computable number is a real number which can be approximated to any arbitrary degree of accuracy by an algorithm. Turing proved that almost all numbers are uncomputable. Chaitin's constant is an uncomputable number, even though it is well defined.
What other systems are equivalent to Turing machines?
The languages that are accepted by a Turing machine are exactly those that are generated by formal grammars. The lambda calculus is a way of defining functions. The functions that can be computed in the lambda calculus are exactly those that can be computed by a Turing machine. These three formulations, Turing machines, formal grammars, and the lambda calculus all look very different, and were all developed by different people. Yet they are all equivalent, and have the same problemsolving power. This is generally taken as evidence for the ChurchTuring thesis, which is the claim that our intuitive notion of an algorithm or an effective procedure is captured by the mathematical definition of a Turing machine.
Electronic computers, and even quantum computers, are exactly equivalent to Turing machines, if they have access to an unbounded supply of memory. As a corollary, all implementable programming languages are at best equivalent in power to a Turing machine (in practice, a very few are less powerful). Such languages are said to be Turingcomplete. Systems equivalent to a Turing machine include:
 Turing machine with several tapes
 Turing machine with a 2dimensional "tape" (an infinite number of linear tapes)
 Turing machine with a limited number of states and tape symbols
 States×symbols can be any of 2×18, 3×10, 4×6, 5×5, 7×4, 10×3, 22×2
 Turing machine with 2 symbols, 176 states, always reversing (http://skelet.ludost.net/bb/RTM.htm) the tape symbol
 Turing machine with 2 symbols, with 3 possible symbol changes: 0>0, 0>1 and 1>1 (proposed by Hao Wang)
 Finite state machine with 2 stacks
 Finite state machine with 2 counters
 Formal grammar
 Post system
 Lambda calculus
 Partial recursive functions
 Almost any modern programming language (when given unlimited memory), including:
 Wang tiles
 Recurrent neural network (finiteprecision inputs/outputs/weights, infiniteprecision signals initialized to zero)
 Cellular automaton, including:
 Conway's Game of Life
 Cellular automaton with just 1 dimension, 2 states, 3 cells per neighborhood (e.g. rule 110)
 Nondeterministic Turing machine
 Probabilistic Turing Machine
 Quantum computer
The last three examples use a slightly different definition of accepting a language. They are said to accept a string if any computation accepts (for nondeterministic), or most computations accept (for probabilistic and quantum). Given these definitions, those machines have the same power as a Turing machine for accepting languages.
What problems require more powerful machines?
Sometimes machines are considered that have more power than a Turing machine. For example, an oracle machine uses a black box that can compute some particular function that might not be possible for an ordinary Turing machine. The computational strength of an oracle machine is described by its Turing degree. The theory of real computation deals with machines using infiniteprecision real numbers. Within this theory, it is possible to prove interesting statements such as "the complement of the Mandelbrot set is only partially decidable". For other such powerful machines, see superTuring computation and hypercomputation.
See also
References
 Hartley Rogers, Jr, Theory of Recursive Functions and Effective Computability, MIT Press, 1987, ISBN 0262680521 (paperback)
External links
 Stanford Encyclopedia of Philosophy entry (http://plato.stanford.edu/entries/computability/)cs:Teorie vyčíslitelnosti
de:Berechenbarkeitstheorie es:Teora de la computabilidad fr:Thorie de la calculabilit ja:計算可能性理論 th:ทฤษฎีการคำนวณได้