People in informal language constantly use "exponential" to mean "more than linear." Which drives me completely insane, but I think it's time to just accept it.
Hell I'd be happy if it were used to describe sequences with more than just two values. I hear it constantly used in comparisons of just two quantities where the intended meaning is just "a lot more".
Yes, this gets to me too. You can fit any sequence of n points with a curve described by a polynomial of n-1 degree. therefore, you need at least 4 points to know that the curve is not cubic, and an infinite number to know that it is exponential (though you can usually surmise as much sooner, because many natural phenomena are exponential or logarithmic).
What frustrates me is people saying "exponentially more" when comparing two things, like the Bolt has exponentially more range than the Leaf. I mean, you could fit an exponential curve to those data points, but also logarithmic, so why bother?
There's no hope of curing this in the popular press, but I think technical people should still be careful with the term. I think it's great that someone's called it out on HN.
People often use big-O (asymptotically bounded above) when they really want big-Θ (theta, asymptotically bounded above and below). Strictly by the definition, for instance, binary search is O(e^x), as well as O(x^12), O(x), and O(lg x). However, binary search is Θ(lg x), and not Θ(any of those other functions).
With big theta you strictly have to talk about best and worst cases, because for e.g. insertion sort, what's true of worst case is not true of best case (and vice versa). For big-O, what's true of worst case is also true of best case, so talking about worst case suffices. Big-O is less precise and therefore more useful, because average case usually = worst case anyway.
That's not right at all. Exponential is constant^x.
Also you can put "n *" in front of all of them if you want, but multiplying by a constant is usually omitted because it doesn't affect the shape of the graph.