Gradient-Sensitive Optimization for Convolutional Neural Networks

Figure 1

(a) Parameter iteration process with a small learning rate. (b) Parameter iteration process with a large learning rate.

(a)

(b)