It's customary to state the optimization problem such that you're minimizing a function - hence "local minima". If you want to maximize something, you simply negate the function.