Evaluations of an algorithm for large multivariate optimization

Many simulation-driven engineering and scientific problems require finding an optimum of a function with many variables. Such settings pose a challenge for standard algorithms due to the large search space which in turn can lead to poor final results. Therefore this paper proposes a new simplified approach in which the dimension of the problem is dynamically reduced during the search to formulate a problem of lower size (dimension) which is easier to solve. A main novelty of the algorithm is its simplicity. Numerical experiments show the potential of this approach.