中文摘要：

In this paper, a novel flow control mechanism in cognitive packet network (CPN) based on the improved back propagation (BP) neural network is proposed, considering the flow distribution status predicted by BP neural network when packets are routed. The objective is to increase the capacity of CPN and improve the quality of service (QoS) by achieving flow balance. Besides, considering the slow convergence speed of traditional BP algorithm and the quick change of the flow status in cognitive packet network, an improved BP algorithm with dynamic learning rate is designed in order to achieve a higher convergence speed. The mechanism, which we propose, regards the predicated traffic data as an important factor when packets are routed to implement flow control. By achieving balance, the quality of network can be improved obviously. The simulation results show that the proposed mechanism provides better average time delay and packets loss ratio.

英文摘要：

In this paper, a novel flow control mechanism in cognitive packet network （CPN） based on the im- proved back propagation （BP） neural network is proposed, considering the flow distribution status predicted by BP neural network when packets are routed. The objective is to increase the capacity of CPN and improve the quality of service （QoS） by achieving flow balance. Besides, considering the slow convergence speed of tradi- tional BP algorithm and the quick change of the flow status in cognitive packet network, an improved BP algo- rithm with dynamic learning rate is designed in order to achieve a higher convergence speed. The mechanism, which we propose, regards the predicated traffic data as an important factor when packets are routed to imple- ment flow control. By achieving balance, the quality of network can be improved obviously. The simulation re- sults show that the proposed mechanism provides better average time delay and packets loss ratio.