What is the meaning of putting uniform distribution inside log? See page 5 of this paper (Corentlin et al.)

enter image description here

To make it more clearer, within my knowledge, I think I should put a single value inside log(). I have no idea what is the meaning that put a distribution inside. Does it result in another distribution? It's strange to me because I think the author want to get a single value.

It seems like this is really an easy question but I just can't figure it out, and I keep finding log-uniform on internet, which I believe it's not.
Thanks in advance!


Corentin Tallec, Yann Ollivier. Can Recurrent Neural Network Warp Time? ICLR 2018

  • 1
    $\begingroup$ 1. This is explicitly discussed in the first sentence on page 5. 2. Please give a full reference and some context for the problem being discussed, not just a link. $\endgroup$
    – Glen_b
    May 14, 2018 at 4:53
  • $\begingroup$ @Glen_b I don't see how the first sentence on page 5, "Values of..." would explicitly discuss the meaning of this notation. But I agree a full reference to the paper is appropriate $\endgroup$ May 14, 2018 at 6:11
  • $\begingroup$ Sorry I'm new to this community. I have edited my post, please check. $\endgroup$ May 14, 2018 at 8:17
  • 1
    $\begingroup$ Why do you believe it's not log uniform? $\endgroup$
    – Peter Flom
    May 14, 2018 at 12:13
  • $\begingroup$ @Juho It explains why there's a log in the expression in the question. $\endgroup$
    – Glen_b
    May 14, 2018 at 17:46

1 Answer 1


This is somewhat informal notation, but it cannot have any other meaning than a random variable which is minus the log of a uniform random variable.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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