Global optimization

From Glossary

Jump to: navigation, search

Generally refers to mathematical programming without convexity assumptions, which are NP-hard. In general, there could be a local optimum that is not a global optimum. Some authors use this term to imply the stronger condition there are multiple local optima. Some solution strategies are given as heuristic search methods (including those that guarantee global convergence, such as branch and bound). As a process associated with algorithm design, some regard this simply as attempts to assure convergence to a global optimum (unlike a purely local optimization procedure, like steepest ascent). See the Glover's linearization.

Personal tools