A new approach for improving the convergence performance of global optimization problems

By introducing the concept of simulated annealing into the conjugate gradient algorithm, we propose a stochastic conjugate gradient algorithm which has an increased probability of obtaining a global minimum, and the determination of the weights of the cost function becomes easier due to the wider feasible scope of its parameters. We apply the proposed algorithm to an optimal task partitioning and compare the scope of the parameters and the probability of obtaining a global minimum with those of the Boltzmann machine. Simulation results show characteristics in favor of the proposed algorithm. We also present a hardware for the proposed algorithm