Twofold type of backpropagation neural network

Various types of neural networks have been introduced and those have been used in various areas. In some areas, those operate in a very good manner, but in another they don´t. For example, Boltzmann machine and Hopfield network have better estimation and analogy abilities compared with backpropagation neural networks, but they are not accurate and are not easily convergent. On the other hand, backpropagation neural networks are very good at learning various patterns, but are bad at estimation and analogy when they are are under a very noisy condition. This means that if backpropagation neural networks can overcome estimation and analogy limitations, this can cover most of the application areas. So in this paper, an analogy and estimation method has been studied by introducing a twofold type of backpropagation neural network. A very good result has been obtained. And also, a new application field of those theories has appeared