A Rectified Linear Unit is a common activation function in deep neural networks and is often abbreviated as "ReLU". I usually pronounce it as /rel-you/ (with the "e" as in "relative" or "rectified"), but I often hear colleagues pronounce it /ray-loo/.

Is there a consensus on the correct pronunciation of "ReLU"?

  • 1
    However the guy who invented it pronounces it. If I had to guess, though, I'd say "ree-loo". It's not "rel-you" because the first syllable is obviously "Re", and it doesn't include the "L". And "Re" by itself would be pronounced "ree", as in "repair" or "refund". – Hot Licks Apr 13 '18 at 0:48
  • 1
    I pronounce it 'reh-loo'. – Mitch Apr 13 '18 at 2:58
  • 1
    @HotLicks I have to disagree with the guy who invented it pronouncing it right (at least they don't some times, don't know about this case) gifs are definitely gifs, not jifs, however I would pronounce it as Mitch does – user300397 May 25 '18 at 21:21

I've always heard professors in classes I've taken at Stanford, such as CS230 and CS231N, pronounce it as /ray-loo/ and /reh-loo/, and I've always used /ray-loo/.

However, the lecturer in this video uses the /rel-you/ pronunciation, which prompted the search that brought me here. The lecturer clearly has extensive ML experience.

I'd hazard a guess based on my anecdotes that /ray-loo/ is more common, but I'd definitely say all three seem to be accepted, given the involvement all parties have in the ML community.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.