Does the word "efficient" usually refer to polynomial time or polylogarithmic time?

Question

This question is strictly about terminology.

I'm not an expert in CS, but I've almost always seen the word "efficient" applied to an algorithm to mean "of polynomial runtime". E.g. this question and the Wikipedia article on computational complexity theory use the word in that way.

However, other sources (e.g. the Wikipedia articles on NC and P-complete) reserve the word "efficient" for algorithms with polylogarithmic runtimes and use "tractable" for algorithms with polynomial runtimes.

This conflicting terminology is obviously confusing. Of course, any (good) formal discussion of algorithmic complexity will precisely define its terminology, but in general, are either of the two usages above clearly more conventional than the other? If someone uses the term "efficient" in an informal discussion of computational complexity without defining it, is it more likely that they mean "polynomial time" or "polylogarithmic time"?

Answer 1

There is no one true answer. It depends on context. The most common context is one where polynomial-time is taken as more or less synonymous with efficient, so if you had no further context, I would certainly guess "polynomial time". Polylogarithmic time is used only in very narrow contexts.

In general, if you think your audience might not be sure about the meaning, then I recommend you to define your terms.

Answer 2

Perhaps this is a reference to Cobham's thesis https://en.wikipedia.org/wiki/Cobham%27s_thesis?wprov=sfla1. Usually the word used is "feasible" though.