Definitions by WonderingNeuron

Karpathy Constant 

The Karpathy constant is one of the best learning rates for the popular Adam (deep neural network) optimizer. It is defined as η = 3e-4. The actual symbol for the constant is α_k.

What is the correct learning rate for adam in this case?

Just use the Karpathy constant dude

Karpathy Constant by WonderingNeuron November 23, 2017