You can find more information in the original paper https://arxiv.org/pdf/1603.00788.pdf basically you want the approximation parameter also on the real line so that you wont have a problem of a too large learning rate will push the sd invalid
You can find more information in the original paper https://arxiv.org/pdf/1603.00788.pdf basically you want the approximation parameter also on the real line so that you wont have a problem of a too large learning rate will push the sd invalid