DE with A new population initialization method for the dynamic multidimensional knapsack problems
Qi Qiu, Junyi He, Tao Zhu · 2022
Differential Evolution (DE) is a well-known continuous optimization algorithm that performs very well on many continuous optimization problems, and some recent studies show that the DE algorithm is experimentally proven to be equally good at solving discrete problems like the Multidimensional knapsack Problem (MKP). But due to the limitation of solving speed, it is difficult for the DE algorithm to perform well in the speed- hungry dynamic multidimensional knapsack problem(DMKP). To this end, this paper proposes a new population initialization method (NPIM), which enables the initial population quality to be improved by leaps and bounds through information reorganization. NPIM can even be used as a separate algorithm to solve MKP when the problem scale is small. Experimentally, it was found that NPIM doubled the speed of the DE algorithm to solve MKP