A State Space Search Algorithm and its Application to Learn the Short-Term Foreign Exchange Rates

We propose the use of a state space search algorithm of the discretetime recurrent neural network to learn the short-term foreign exchange rates. By searching in the neighborhood of the target trajectory in the state space, the algorithm performs nonlinear optimization learning process to provide the best feasible solution for the nonlinear least square problem. The convergence analysis shows that the convergence of the algorithm to the desired solution is guaranteed. The stability properties of the algorithm are also discussed. The empirical results show that our method is simple and effectively in learning the short-term foreign exchange rates and is applicable to other applications.