Comparative Evaluation of Generalized ADALINE using Variable Learning Rate Parameter

The ADAptive LINear Element (ADALINE) neural network uses Least Mean Square (LMS) learning rule. This paper presents a comprehensive comparison of three different variable learning rate (VLR) parameter LMS algorithms, for the generalized ADALINE neural network paradigms. These algorithms are used to adjust the weights of the ADALINE neural network, which are tested under three different applications: adaptive prediction, system identification and noise cancellation. The performance analysis of the algorithms for different scenarios is discussed with the help of computer simulations. The simulation results show that in the initial stage, the mean square error between the network output and the desired output is less for the algorithm that has the fastest convergence rate. Therefore, the major advantage is faster convergence of synaptic weights towards the optimum solution, in addition to the better tracking performance in different applications.