Incremental learning for SIRMs fuzzy systems by Adam method
Shu Matsumura, Tomoharu Nakashima · 2017
This paper proposes an incremental approach for training SIRMs fuzzy systems. In the domain incremental learning, a single training example is used to modify the model parameters during the training phase. This paper suggests to use the Adam method, which has the advantage of obtaining stable results by automatically adapting the learning rates in the training of model parameters. Computational experiments are conducted to investigate classification performance as well as modeling performance of SIRMs fuzzy systems trained by Adam method. The experimental results show the advantage of the proposed method over the common stochastic gradient descent algorithm.