Comment by tavianator

Presumably that's just a mistake. The author calls it "stochastic gradient descent" correctly elsewhere in the article