Tree learning theory of neural networks

A tree learning theory more similar to human learning is introduced. In this paper learning is defined as not only changes of weights but also changes of definite relationships between the informations of the outward world and the firing states of a neural network, which is constructed by three kinds of neural groups called sense, knowing, and reasoning groups respectively. The learning algorithms of the first two are completed. It has been proved that the learning algorithms can perform cognition tasks of objects from informations of the outward world more reasonably and faster.