Which of the following are Pros of SGD?

• A. Stochastic Gradient Descent (SGD) requires several hyperparameters
• B. It is sensitive to feature scaling
• C. Stochastic Gradient Descent (SGD) is very efficient
• D. All of the above