In this paper, the simultaneous perturbation stochastic approximation （SPSA） algorithm is used for seeking optimal parameters in an adaptive filter developed for assimilating observations in the very high dimensiona...In this paper, the simultaneous perturbation stochastic approximation （SPSA） algorithm is used for seeking optimal parameters in an adaptive filter developed for assimilating observations in the very high dimensional dynamical systems. The main results show that the SPSA is capable of yielding the high filter performance similar to that produced by classical optimization algorithms, with better performance for non-linear filtering problems as more and more observations are assimilated. The advantage of the SPSA is that at each iteration it requires only two measurements of the objective function to approximate the gradient vector regardless of the dimension of the control vector （or maximally, three measurements if second-order optimization algorithms are used）. The SPSA approach is thus free from the need to develop a discrete adjoint of tangent linear model as it is required up to now for solving optimization problems in very high dimensional systems. This technique offers promising perspectives on developing optimal assimilation systems encountered in the field of data assimilation in meteorology and oceanography.展开更多
文摘In this paper, the simultaneous perturbation stochastic approximation （SPSA） algorithm is used for seeking optimal parameters in an adaptive filter developed for assimilating observations in the very high dimensional dynamical systems. The main results show that the SPSA is capable of yielding the high filter performance similar to that produced by classical optimization algorithms, with better performance for non-linear filtering problems as more and more observations are assimilated. The advantage of the SPSA is that at each iteration it requires only two measurements of the objective function to approximate the gradient vector regardless of the dimension of the control vector （or maximally, three measurements if second-order optimization algorithms are used）. The SPSA approach is thus free from the need to develop a discrete adjoint of tangent linear model as it is required up to now for solving optimization problems in very high dimensional systems. This technique offers promising perspectives on developing optimal assimilation systems encountered in the field of data assimilation in meteorology and oceanography.
