issues
search
microsoft
/
prv_accountant
A fast algorithm to optimally compose privacy guarantees of differentially private (DP) mechanisms to arbitrary accuracy.
MIT License
65
stars
10
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
Add computation of trade-off curves
#39
wulu473
opened
8 months ago
0
fix: determine points for mean computation based on inputs
#38
Solosneros
closed
1 year ago
0
Discrete mean differs from continuous mean significantly
#36
xuefeng-xu
closed
1 year ago
10
docs: update DP-SGD example
#35
Solosneros
closed
1 year ago
0
Bump version
#33
wulu473
closed
2 years ago
0
Update readme
#32
wulu473
closed
2 years ago
0
Add composition of approximate DP guarantees
#31
wulu473
closed
2 years ago
0
Add Gaussian and pure DP mechanism
#30
wulu473
closed
2 years ago
0
Add laplace mechanism
#29
wulu473
closed
2 years ago
0
Add convenience routines for DP-SGD
#28
wulu473
closed
2 years ago
0
Cannot import scipy.fft for scipy 1.3.3 and numpy 1.18.1
#27
wulu473
opened
2 years ago
0
Don't exit 0 with linter errors
#26
wulu473
closed
2 years ago
0
Refactor `Composer` interface to allow for heterogeneous composition
#25
wulu473
closed
2 years ago
0
Fix github tag to point to latest release
#24
wulu473
closed
2 years ago
0
Bump version
#23
wulu473
closed
2 years ago
1
Add a strict `--fail-on-no-value` option to the compute epsilon script
#22
wulu473
closed
2 years ago
0
Add a test to check that not more compositions than max_compositions is requested
#21
wulu473
closed
2 years ago
0
Add assertion to make domain creation meets requirements
#20
wulu473
closed
2 years ago
0
Fix typo in test prvs names
#19
wulu473
closed
2 years ago
0
Add check to prevent small values of delta
#18
wulu473
closed
2 years ago
0
PRV accountant breaks for small delta values
#17
mshubhankar
opened
2 years ago
5
Improved conversion from RDP to Approximate DP
#16
s-zanella
closed
2 years ago
0
Add heterogenous composition
#15
wulu473
closed
2 years ago
0
Numerical integration to compute mean fails for some PRVs
#14
wulu473
closed
2 years ago
0
Move RDP computation to PRV and tidy up dependencies
#13
wulu473
closed
3 years ago
0
Improve robustness
#12
wulu473
closed
2 years ago
1
Update readme
#11
wulu473
closed
3 years ago
0
Add workflows for publishing to pypi
#10
wulu473
closed
3 years ago
0
Add workflow for windows
#9
wulu473
closed
3 years ago
0
Fix README example
#8
wulu473
closed
3 years ago
0
Update example in README to use max_compositions argument
#7
wulu473
closed
3 years ago
0
Use longdouble instead of float128 which automatically falls back to double if float128 is not available
#6
wulu473
closed
3 years ago
0
Refactor epsilon computation
#5
wulu473
closed
3 years ago
0
Handling the case without subsampling
#4
YinTat
closed
3 years ago
0
Fix floating point warnings from numpy
#3
wulu473
closed
3 years ago
0
Use more suitable argument names
#2
wulu473
closed
3 years ago
0
Initial PR
#1
wulu473
closed
3 years ago
0