A hybridmeta-heuristicforglobaloptimisationusinglow-discrepancy sequences ofpoints

A hybridnovelmeta-heuristictechniqueforbound-constrainedglobaloptimisation(GO)isproposed in thispaper.Wehavedevelopedaniterativealgorithmcalled LP Optimisation (LP O) thatuseslow- discrepancy sequencesofpointsandmeta-heuristicknowledgetofindregionsofattractionwhensearch- ing foraglobalminimumofanobjectivefunction.Subsequently,thewell-knownNelder–Mead(NM) simplex localsearchisusedtorefinethesolutionfoundbythe LP O method. Thecombinationofthetwo techniques (LP O and NM) providesapowerfulhybridoptimisationtechnique,whichwecall LP NM. Its properties—applicability, convergence,consistencyandstabilityarediscussedhereindetail.The LP NM is testedonanumberofbenchmarkmultimodalmathematicalfunctionsfrom2to20dimensionsand compared withresultsfromotherstochasticheuristicmethods.