Explaining how the Learning Rate hyperparameter starts high (impactful) during early training stages, but decreases during each subsequent learning stage.