摘要：AbstractAn important requirement in control systems is an acceptable transient response. When the underlying dynamical systems are unknown, a factor that contributes is how fast an algorithm can identify them. It is well-known that the back propagation algorithm has rather poor convergence properties. Consequently, it sometimes takes several thousand iterations before the transient response is satisfactory. In this paper, we show that a sequential variant of the extreme learning machine considerably improves the transient response.
