Tree annealing for constrained optimization

In this article, we introduce an optimization algorithm that integrates the basic idea of simulated annealing and partitioning of the search domain by binary tree of subdomains. Also included in the algorithm are procedures that implement direct simple constraints on the decision variables. The algorithm can be used to solve various optimization problems arising from parameter identification, neural network training, and nonlinear least squares. Results on several examples are presented to illustrate typical performance of this algorithm after its presentation.