[home] [lexicon] [problems] [tests] [courses] [auxiliaries] [notes] [staff] | ||

Mathematics-Online lexicon: | ||

## Steepest Descent |

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z | overview |

The method of steepest descent is used to minimize multivariate functions . In one step of the iteration, first, the negative gradient

As shown in the figure, the search direction is orthogonal to the level set through and touches the level set for a smaller function value in .

The convergence of the sequence , generated by the the method of steepest descent, can be shown under fairly general assumptions. It is sufficient that is bounded from below and is Lipschitz continuous in a neighborhood of the set , i.e.,

To ensure convergence of the algorithm, it is not necessary to find the exact one-dimensional minimium in the direction of . It is sufficient to use a non-optimal decent direction and obtain merely a reduction of the current function value which is proportional to .

automatically generated 6/14/2016 |