SPAN: a neural network that grows

Summary form only given, as follows. A novel neural model called SPAN (space partition network), which allows the network to adapt its structure autonomously, is proposed. According to the evolution level operators and their selection rules to be defined, a network can generate new neurons and connect them on desired positions in the network when the processing capability of the network is not sufficient, or the network can annihilate nonactive existent neurons. Using this algorithm, the authors initially put a set of seed neurons in the network, then let the network grow according to the training patterns. It is observed from the simulation results that the network will eventually grow to such a configuration which is suitable to the class of problems characterized by the training patterns, i.e. the neural network self-synthesizes itself to fit the problem space.<>