Your question is LSTM Gates and Gradients. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
How do you handle vanishing and exploding gradients in deep recurrent architectures, and what is the mathematical intuition behind LSTM gates?