Search space reduction approach in evolutionary algorithms: The case of high-dimensional portfolio replication problem

It is difficult to find optimum solutions by using evolutionary algorithms (EAs) in high-dimensional search space as compared with low-dimensional search space. For solving such a high-dimensional problem, we propose a search space reduction approach in EAs. For reducing the search space, the approach repeats fixing the value of one design variable which should be reduced from the search space. We apply this approach to optimize the high-dimensional portfolio replication problem. In the numerical experiments, we show that our search space reduction approach in EAs finds good solutions in the reduced search space as compared with a traditional approach which searches solutions in the whole solution space.