Stochastic gradient descent - AshokBhat/ml GitHub Wiki

Stochastic Gradient Descent (SGD)

  • A type of gradient descent

FAQ

  • How is it different from other forms of gradient descent?
    • Using a batch size of 1 random sample.
  • What is the advantage?
    • Computationally less intensive.

-

Does it work?

  • Yes, it does work. The path to minima will be staggered.

See also