Hi, I have added a scheduling plan which I found in "Practical Recommendations for Gradient-Based Training of Deep Architectures" by Yoshua Bengio, Llink to paper.
It keeps the quantity constant for a given amount of epochs and then starts to decrease by 1/epoch_nr.
I have added a test case too.
Hi, I have added a scheduling plan which I found in "Practical Recommendations for Gradient-Based Training of Deep Architectures" by Yoshua Bengio, Llink to paper.
It keeps the quantity constant for a given amount of epochs and then starts to decrease by 1/epoch_nr. I have added a test case too.