Neuron-based connection acceptance strategy

A neural system with the back propagation learning algorithm is discussed for making decisions in connection management. The strategy is divided into two developed stages. In the learning stage, we collect three groups of information including the traffic description, QoS requirement, and link status, and then embed them in the input pool. The accept/reject decision for a requested service is generated in the output layer. Upon completing the learning stage, the on-line test is in progress. From the results, we found that the strategy can support the real-time feature in connection management. In addition, the QoS is guaranteed while the connection is permitted under our ATM testbed.