Learning rule with fractional-order average momentum based on Tustin generating function for convolution neural networks
Jing Jian, Zhe Gao, Tao Kan · 2020 IEEE 9th Data Driven Control and Learning Systems Conference (DDCLS) · 2020
In this paper, we propose a fractional-order average momentum (FOAM) method based on Tustin generating function to train parameters in convolution neural networks. Taking the classical data set MNIST as the training and testing data, the effectiveness of the FOAM for CNNs is verified. The experimental results show that the stochastic gradient descent method based on FOAM can improve the recognition accuracy and learning convergence speed of convolution neural networks.