Multi-objective Optimization Based On Lamarckian Learning With Application

Nowadays, the research on multi-objective optimization algorithm has been playing an important role in the area of evolutionary computation. Lamarckian learning theory provides a new idea to solve the problem from the theoretical level of cultural evolution. Lamarckian learning has been introduced into evolutionary computation to enhance the ability of local search, which gradually develops into a new hot spot-memetic computation, a new path for solving multi-objective optimization problem.This paper first reviews the correlative background of multi-objective optimization. Subsequently, we focus on concepts and classic algorithms of multi-objective optimization. On the basis of above, we describe Nondominated Neighbor Immune Algorithm (NNIA) and Lamarckian theory.In chapter three, Lamarckian learning and dynamical niching are introduced into NNIA. A novel multi-objective optimization algorithm, Nondominated Neighbor Immune Algorithm based on Lamarckian learning (LNNIA), is proposed. The mode, which local search is added in the latter part of running, can accelerate convergence of the algorithm while reducing the number of evaluations. The introduction of niche technology improves diversity of population.In chapter four, Lamarckian learning and Tchebycheff approach are introduced into NNIA. A novel hybrid multi-objective optimization algorithm, Multi-objective Lamarckian Immune Algorithm (MLIA), is proposed. The location of local search added, which is after proportional cloning of NNIA, not only inherits the advantages of the original algorithm, but also compensates for the shortcomings. The introduction of Tchebycheff approach raises the efficiency of local search.Finally, we introduce the algorithm frame in chaper four into multicast routing problem. And the results demonstrate that the new method can solve the multicast routing problem successfully compared with traditional algorithms.