Abstract
An optimization algorithm is presented which may be considered as a variant of Simulated Annealing with an ensemble of K interacting Systems. This algorithm avoids the explicit specification of a temperature parameter and a cooling schedule. The Boltzmann distribution of energies or cost values is accomplished by enforcing an approximately constant total energy of the ensemble during a certain sequence of steps. The implicit temperature is lowered by a cooling step which is controlled solely by the state of the ensemble as a whole. To compare this algorithm with classical Simulated Annealing a problem with known optimal solution is studied with both algorithms. In particular the success probability P and the number of function evaluations required for reaching the optimal solution are compared.