Adaptive learning algorithm and its convergence analysis with complex-valued error loss network