Generalization and comparison of Alopex learning algorithm and random optimization method for neural networks

For the minimum description of length, a penalty term is added to the cost function to reduce the network´s complexity. Alopex learning and random optimization method based neural networks are investigated. The Alopex algorithm is a stochastic learning algorithm for training neural networks of any topology, including feedback loops. The neurons are not restricted to any transfer function and the learning can use any error norm measure. The random optimization method by Matyas (1965) and its modified algorithm are studied and compared with the Alopex algorithm to some adaptive control problems. Simulation results show the pros and cons between two