vc1492a opened this issue 4 years ago · 0 comments
We use a learning rate to help minimize over fitting but the parameters need to be tuned after balancing the data and selecting an architecture.