4634
правки
Glk (обсуждение | вклад) (Новая страница: «'''Approximation algorithm''' --- аппроксимирующий алгоритм. For the ''travelling salesman problem'', as indeed for any other ''intractabl…») |
KEV (обсуждение | вклад) Нет описания правки |
||
Строка 1: | Строка 1: | ||
'''Approximation algorithm''' | '''Approximation algorithm''' — ''[[аппроксимирующий алгоритм]].'' | ||
For the ''travelling salesman problem'', as indeed for any other | For the ''travelling salesman problem'', as indeed for any other | ||
''intractable problem'' it is useful to have a polynomial time algorithm which will | ''intractable problem'' it is useful to have a polynomial time [[algorithm]] which will | ||
produce, within known bounds, an approximation to the required result. | produce, within known bounds, an approximation to the required result. | ||
Such algorithms are called '''approximation algorithms'''. Let <math>L</math> be | Such algorithms are called '''approximation algorithms'''. Let <math>\,L</math> be | ||
the value obtained (for example, this may be the length of a travelling salesman's | the value obtained (for example, this may be the length of a travelling salesman's | ||
circuit) by an approximation algorithm and let <math>L_{0}</math> | circuit) by an approximation algorithm and let <math>\,L_{0}</math> | ||
be an exact value. We require a performance guarantee for | be an exact value. We require a performance guarantee for | ||
the approximation algorithm which could, | the approximation algorithm which could, for a minimisation problem, be stated in | ||
the form: | the form: | ||
Строка 14: | Строка 14: | ||
For a maximisation problem | For a maximisation problem | ||
we invert the ratio <math>L/L_{0}</math>. Of course, we would like <math>\alpha</math> to be | we invert the ratio <math>\,L/L_{0}</math>. Of course, we would like <math>\,\alpha</math> to be | ||
as close to one as possible. | as close to one as possible. | ||
Unfortunately, not every heuristic produces a useful approximation algorithm. | Unfortunately, not every heuristic produces a useful approximation algorithm. |