Draw the Learning Rate against loss curve to detect the fastest decreasing gradient, which is not the minimal point in this line.
Once it reaches the best learning rate, the iteration stops.
Same idea applies for number of epochs
https://arxiv.org/pdf/1506.01186.pdf
https://blog.csdn.net/guojingjuan/article/details/53200776
https://www.jeremyjordan.me/nn-learning-rate/
https://medium.com/@lipeng2/cyclical-learning-rates-for-training-neural-networks-4de755927d46
https://github.com/bckenstler/CLR
jupyter shoortcut
command completion: TAB
parameter completion: shift+TAB
command doc: ?
source code: ??
H: show all shortcuts in main screen
网友评论