论文部分内容阅读
A Neural Network( NN ) approach to ABR flow control algorithm in ATM networks is proposed. The NN predicts the queue length, its variation and possible cell loss, then regulates the source rate adaptively. Therefore, more appropriate value of the explicit rate can be determined in the corresponding field of the RM cells. This approach performs better than the traditional static feedback control. Additionally, the performance of this algorithm under CBR background traffic is discussed, and the simulatino results show that the neural network is also efficient.
A NN (NN) approach to ABR flow control algorithm in ATM networks is proposed. The NN predicts the queue length, its variation and possible cell loss, then regulates the source rate adaptively. Thus, more appropriate value of the explicit rate can be determined in the corresponding field of the RM cells. This approach performs better than the traditional static feedback control. The performance of this algorithm under CBR background traffic is discussed, and the simulatino results show that the neural network is also efficient.