[Retracted] Gradient Descent Optimization in Deep Learning Model Training Based on Multistage and Method Combination Strategy

Table 2

The differences between multistage strategy and CLR/SGDR.

Features	CLR and SGDR	Proposed multistage strategy
The length of timepieces	Fixed or predefined	Changeable
Methods for each timepiece	Only predefined	Predefined but changeable