Neural Networks (NN) have been taken notice by researchers in the world since 1950's. It is believed among people who study computer science that they will be able to process informations flexibly and speedy.
In this study, three kinds of NNs which were created, play a game to prove the properties. In the game, a player chases some targets in a square field. Players are operated by each NN. The NNs learn appropriate movement (supervising signal) by the method of Back Propagation.
One of the three NNs, which is the simplest, was fed the signal that was preceding-ly well processed environmental information. The second NN was fed 100 inputs that were binary signals. These two could learn enough to play the game. The third one was fed 100 inputs that were continuous-value, which did not work. The simplest NN is the best learning to play the game. The second NN could not learn just supervising signal. But it was good enough to play and it could learn output near right against unknown inputs.
The results of the game are not enough to prove the NNs' flexibility and speed of processing. But I think it shows a little possibility.