Smart neural nets for fast learning

The 'smart neural nets' for fast and well behaved learning are formed by completely avoiding the use of the sigmoid non-linear function and by redesigning the neurons appropriately. Just like the conventional neural nets, smart neural nets are trained using the regular innovation backpropagation learning algorithm. The implementation of the two-input exclusive-OR gate is used to compare the performance of the smart neural nets against the conventional neural nets toward the regular innovation backpropagation learning. It is found that unlike the conventional neural nets, the smart neural nets learn extremely fast and smoothly. In addition, it is shown that the sigmoid non-linear function or hard limiter non-linear function driven conventional neurons are not important to build feed forward neural nets.