A new hybrid genetic algorithm for global optimization

We present a hybrid genetic algorithm for finding guaranteed and reliable solutions of global optimization problems. This algorithm uses the branch-and-bound principle to obtain small regions where candidate solutions lie. In this way, a highly-performing initial population is formed and initial bounds for the global minimum \(f^*\) are obtained. By applying a genetic algorithm using this population as well as a safe and reliable technique for updating properly the bounds of \(f^*\), we are able to compute with certainty global minima for various difficult test problems. The proposed algorithm becomes more effective when a new termination criterion is used. This criterion is based on the notion of a shrinking box and using this extra computational effort is avoided. HIG has exhibited high performance when applied to difficult problems, especially to multimodal and high-dimensional objective functions. It has been clear that HIG outperforms traditional genetic algorithms. Also, for all the problems examined, HIG has given better results for both Algorithm 1 and 2 studied separately.

Recommendations

Cites work

Cited in

(14)

Describes a project that uses

Uses Software

Genocop

This page was built for publication: A new hybrid genetic algorithm for global optimization

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4378896)