Michael Shoemate
Michael Shoemate
MPFR has a much higher upper bound on the exponent than native float types do, at the same level of precision. Because of this, `inf_*` operations may evaluate to floating-point...
Closes #365 This PR contains a minimal R package boilerplate as well as a working end-to-end demonstration of how to call into the opendp rust library from R. Please consult...
We need a combinator that accepts a measurement with a ZeroConcentratedDivergence output measure and emits a measurement with a SmoothedMaxDivergence measure. I'm looking at using the bound described in the...
We want to express this in terms of product constructions: ```rust // defined as d(x, x') = sum_i d_{inner}(x_i, x'_i), where x_i is the ith partition struct ProductMetric { _inner:...
Using [algorithms](https://github.com/opendp/prelim/commits/Shuronglin-StratifiedSampling-CI) from @Shuronglin! [All `alg_Noise*.ipynb` files.](https://github.com/opendp/prelim/tree/Shuronglin-StratifiedSampling-CI/python)
It is not necessary to add significant new machinery to represent postprocessors-- postprocessors can simply be transformations with maps and metrics stripped away, or rather, with `AgnosticMetric` as the input...
We have binary search utilities in Python, but not in Rust. It would be ideal to have symmetry in our interfaces across languages. There is already a PR for this,...
Similar in nature to the reconstruction and membership attack notebooks, it would be useful to have a notebook demonstrating a record linkage attack. As in, show how individuals in a...
We should add constructors for aggregation transformations that use modular arithmetic, as well as noise addition mechanisms that add modular nose. Our current `sized_bounded_sum` uses a constructor check to ensure...
If we have an exponential mechanism that takes in a vector of scores (#324), we could add an aggregator for quantile utilities. The transformation could behave as follows-- ``` input_domain:...