Changes between Initial Version and Version 1 of org/meetings/2018Rome/Notes/FloatingPointNoise


Ignore:
Timestamp:
Mar 15, 2018, 1:12:18 PM (2 years ago)
Author:
teor
Comment:

Initial notes

Legend:

Unmodified
Added
Removed
Modified
  • org/meetings/2018Rome/Notes/FloatingPointNoise

    v1 v1  
     1== Random Floating Point ==
     2
     3Generate uniformly random float with probability 1/2 in [0.5, 1.0), probability 1/4 in [0.25, 0.5), and so on…
     4
     5== Differential Privacy with Floating Point ==
     6
     7Clamp, then add noise, then snap/bin, then clamp.
     8
     9Randomising the low bits after scaling doesn't necessarily work: snapping correctly hides the low bits.
     10
     11== Floating Point Distributions ==
     12
     13Error of sqrt(), cos(), and ln().
     14
     15Do a floating-point error analysis on the alternative Gaussian transforms, and then pick one.
     16
     17Use an argument similar to the one at the end of section 5.2 in:
     18https://www.microsoft.com/en-us/research/wp-content/uploads/2012/10/lsbs.pdf
     19
     20Use a numerical methods text for reference.