Then to optimize over that linear model they’ll show you gradient descent, and again you’ll note, “Well, that’s just calculus!” Support vector machines? Merely convex optimization. Naive Bayes? Just ...