| 摘要 |
PURPOSE: To always improve or maintain assigning performance by conducting neural-net re-learning all the time, returning it to assignment with the original net if performance degrades after learning, replacing the original net, and conducting up-to-date storage of the net if performance is improved, if perfor mance is improved. CONSTITUTION: This controller is provided with a learning means 13 for conducting re-learning of a neutral net 12 for assignment as necessary. After learning, an assignment, performance evaluation means 15 compares assigning performance before and after learning with each other. A storage means 16 maintain the neural-net before learning, left as it is if performance after learning is worse than that before learning, and if better, up-to-date storage of the neural- net 12 after learning is conducted as a succeeding neural-net 12 for allocation. |
