Also known as: SGD
SGD (Stochastic Gradient Descent) is a gradient descent method that updates model parameters using one randomly selected training example or a very small sample at a time. It is computationally efficient and introduces noise that can help optimization, but may require careful learning-rate control.
A gradient descent algorithm in which the batch size is one. In other words, SGD trains on a single example chosen uniformly at random from a training set. See Linear regression: Hyperparameters in Machine Learning Crash Course for more information.