A new approach to global minimum and its applications in blind equalization

A new approach to the global minimum of the cost function of a BP neural network is proposed in the paper, which combines the merits of the Rosario algorithm and the random optimization method. Its cost function has strict convex character (after a threshold) and converges much faster than the conventional backpropagation method. As an example, we evaluated its performance by using it in blind equalization. With the help of higher-order cumulants (HOC), the novel blind equalization scheme converges much faster than the CMA (constant modulus algorithm) algorithm and is superior to the equalizer using the conventional backpropagation method due to its ability of finding the optimal solution with relatively fewer iteration steps