DOA Estimation Using Neural Network with Levenberg-Marquardt Training

Direction of arrival (DOA) is a classic problem in the field of array processing and communications. In this work, we investigate the efficiency of Levenberg-Marquardt algorithm for neural network training in solving DOA problem. Numerical simulations have been carried out to make a comparison between the performance of the ordinary multi-layer perceptron (MLP) and MLP network with Levenberg-Marquardt training. The simulation results show that network with Levenberg-Marquardt training and fewer numbers of neurons outperforms MLP with gradient descent training in terms of mean square error (MSE). In fact, the ordinary MLP network with gradient descent training has failed to provide acceptable estimates for DOAs.

کلیدواژه ها:

Levenberg-Marquardt ، Direction of Arrival ، Multi-layer percepteron