An Immune System-Based Genetic Algorithm to Deal with Dynamic Environments: Diversity and Memory

Abstract

The standard Genetic Algorithm has several limitations when dealingwith dynamic environments. The most harmful limitation as to do withthe tendency for the large majority of the members of a populationto convergence prematurely to a particular region of the search space,making thus difficult for the GA to find other solutions when changesin the environment occur. Several approaches have been tested toovercome this limitation by introducing diversity in the population(hypermutation, random immigrants or genetic operators) or throughthe incorporation of memory in order to help the algorithm when situationsof the past can be observed in future situations (diploidy, case-basedmemory). In this paper, we propose a GA inspired in the immune systemideas in order to deal with dynamic environments. This algorithmcombines the two aspects mentioned above: diversity and memory. Withrespect to diversity, we use a biological inspired mechanism thatemulates the process of somatic hypermutation that is present duringclonal selection. Concerning memory, our GA relies upon a sort ofmemory mechanism inspired by the memory B-cells of the immune system.We empirically study our approach with a standard combinatorial problemas test bed, namely the dynamic version of the 0/1 knapsack problem,which positively compares with other proposals. In particular, wewill show that our algorithm improves its response over time, a kindof secondary response present in the immune system, e.g. it is capableof learning. Moreover, we will show that our algorithm is also moreadaptable and accurate than the other algorithms proposed in theliterature.