Title :
Managing buffer effectively by the back propagation learning algorithm
Author :
Wu, Yuanming ; Gao, Ke
Author_Institution :
Sch. of Optoelectron. Inf., Univ. of Electron. Sci. & Technol. of China, Chengdu, China
Abstract :
A novel buffer management algorithm named dynamic neural sharing (DNS) is suggested in this paper. It is based on the back propagation learning algorithm (BPA). Simulations show that it can reduce the packet loss in self-similar traffic patterns, and a trade-off between packet loss and fairness can be realized well than those traditional algorithms such as complete partitioning (CP), complete sharing (CS) and static partial sharing (SPS).
Keywords :
backpropagation; neural nets; telecommunication network management; back propagation learning algorithm; buffer management algorithm; complete partitioning; complete sharing; dynamic neural sharing; self-similar traffic patterns; static partial sharing; Autocorrelation; Channel allocation; Management training; Neural networks; Packet switching; Partitioning algorithms; Power system modeling; Telecommunication traffic; Time division multiplexing; Traffic control;
Conference_Titel :
Communications and Information Technology, 2005. ISCIT 2005. IEEE International Symposium on
Print_ISBN :
0-7803-9538-7
DOI :
10.1109/ISCIT.2005.1566936