主催: (社)計測自動制御学会システムインテグレーション部門
p. 402
When the environment changes over time, we need to make additional learning in order to adapt the new environment. However, if we make only additional learning, the rule becomes more and more complex which causes poor generalization ability. Then, it is important to decrease the effect of data causing misclassification. This is called forgetting. In this paper, the effectiveness of additional learning and forgetting will be shown along with several examples.