Application of the square contour algorithm in blind equalizers based on complex neural networks
Juan Zhao · 2013
The error function is important for the blind equalizer based on neural networks to adaptively adjust its parameters. Aiming at finding a new error function, the paper studied the square contour algorithm (SCA) and the complex backward propagation neural networks (CBPNN). The properties of the equalizers based on the cost function of SCA were simulated, and comparison was made with that of CMA. Results show that the equalizer with cost function of SCA converges slower and the byte-error rate (BER) is greater than that of CMA. The residual errors are the same because the cost function only varies in appearance. Therefore, in designing the equalizer based on CBPNN, it is not advisable to replace the error function of CMA with that of SCA.