A learning rate of 98% implies an exceptionally rapid rate of improvement.

Question

Quizplus · Accepted Answer

A learning rate of 98% implies that the model changes its parameters by 98% of the gradient value at each iteration. This is a very high learning rate and can lead to overshooting of the minimum value and unstable behavior of the model. Therefore, it does not necessarily imply an exceptionally rapid rate of improvement. A smaller learning rate may be better suited for stable and consistent improvement of the model.

A Learning Rate of 98% Implies an Exceptionally Rapid Rate