Call admission control (CAC) is one of the most fundamental preventive congestion control mechanisms in asynchronous transfer mode (ATM) networks. Call admission control is further complicated by the diverse mix of traffic types and service requirements and by the increasing speed of transmission. In this project, the focus is on a call admission control with interval arithmetic coulomb energy network, a generalization of coulomb energy network (CEN), which can handle interval data and point data for the learning efficiency. The approximations are based on the continuous monitoring of the network state and the use of a feed forward multilayer perceptron. The call admission control decision whether to accept the connection or refuse it based on the current load conditions is hence verified. The performance of the neural network based on different learning rates and the improved numerical efficiency is illustrated using three example activation functions, the sigmoidal function, bipolar and the hyperbolic tangent functions. The experimental evidence for the analysis of a given network shows that the performance of the network can be enhanced when lesser learning rates are used but with minor acceptable variations in the delay as compared with greater learning rates. The delay comparison using learning rates of 0.05 and 0.01 is made with the sigmoidal, bipolar and the hyperbolic tangent activation functions.Keywords: CAC, ATM, CEN, activation function
