Mini-batch gradient descent: A variant of gradient descent that updates the parameters of the model using a small fixed-size subset of the training data instead of the entire dataset or a single sample.