Chaotic optimization method combined with coordinate transformation

We propose a nonlinear optimization method based on a combination of chaotic dynamics and a coordinate transformation. The coordinate transformation serves to make small eigenvalues of the curvature matrix of the objective function, which is defined for each optimization problem. Then, the objective function becomes flat so that the chaotic dynamics is able to search for possible solutions over a wide domain. We apply our method to a chaotic version of the doubly constrained network model (DCN) that solves quadratic assignment problems. As a result, our new method achieves solution improvements.