Changes between Version 1 and Version 2 of Ticket #23061, comment 69


Ignore:
Timestamp:
Oct 25, 2018, 3:47:41 AM (14 months ago)
Author:
riastradh
Comment:

Legend:

Unmodified
Added
Removed
Modified
  • Ticket #23061, comment 69

    v1 v2  
    1 If you take the uniform distribution on ^^[0,1^^] or (0,1), i.e. the Lebesgue measure, and round it to the nearest floating-point number, the distribution on the result is the natural notion of the uniform distribution on floating-point numbers in ^^[0,1^^].
     1If you take the uniform distribution on ^^[0,1^^] or (0,1), i.e. the Lebesgue measure, and round it to the nearest floating-point number, the distribution on the result is the natural notion of the uniform distribution on floating-point numbers in ^^[0,1^^]: the probability of each floating-point number is the width of the interval of real numbers rounded to that floating-point number.
    22
    33Note that even if you round a sample from the open interval (0,1) of real numbers, the result might be rounded to 0 or 1, if it is in (0, 2^emin - p - 1^] or [1 - eps/4, 1):