Heuristic random optimizer-version II

This work presents the search logic of a multivariable optimizer that is a combination of heuristic rules and random perturbations, and demonstrates its performance on classic and new test cases. One of the neural network applications had 80 decision variables. The improved Heuristic Random Optimization (HRO-II) strategy is shown to accommodate surface discontinuities, handle hard constraints, to have a high probability of finding a global optimum, and to be scale independent.