Mode-Utilizing Developmental Learning Based on Coherent Neural Networks

We propose a mode-utilizing developmental learning method. Thereby a system possesses a mode parameter and learns similar or advanced tasks incrementally by using its cumulative skill. We construct the system based on the coherent neural network where we choose its carrier frequency as the mode parameter. In this demonstration, we assume two tasks: basic and advanced. The first is to ride a bicycle as long as the system can before it falls. The second is to ride as far as possible. It is demonstrated that the system finds self-organizingly a suitable value of the mode parameter in the second task learning. The learning is performed efficiently to succeed in riding for a long distance.