Find the highest-utility (d_in, d_out)-close Transformation or Measurement.

Searches for the numeric parameter to make_chain that results in a computation that most tightly satisfies d_out when datasets differ by at most d_in, then returns the Transformation or Measurement corresponding to said parameter.

Usage

binary_search_chain(make_chain, d_in, d_out, bounds = NULL, .T = NULL)

Arguments

make_chain: a function that takes a number and returns a Transformation or Measurement
d_in: how far apart input datasets can be
d_out: how far apart output datasets or distributions can be
bounds: a 2-tuple of the lower and upper bounds on the input of make_chain
.T: type of argument to make_chain, either "float" or "int"

Value

a Transformation or Measurement (chain) that is (d_in, d_out)-close.

Details

See binary_search_param to retrieve the discovered parameter instead of the complete computation chain.

Examples

enable_features("contrib")
# create a sum transformation over the space of float vectors
s_vec <- c(vector_domain(atom_domain(.T = "float")), symmetric_distance())
#> Error in .Call("domains__atom_domain", bounds, nullable, .T, rt_parse(.T.bounds),     log, PACKAGE = "opendp"): "domains__atom_domain" not available for .Call() for package "opendp"
t_sum <- s_vec |> then_clamp(c(0., 1.)) |> then_sum()
#> Error: object 's_vec' not found

# find a measurement that satisfies epsilon = 1 when datasets differ by at most one record
m_sum <- binary_search_chain(\(s) t_sum |> then_laplace(s), d_in = 1L, d_out = 1.)
#> Error in exponential_bounds_search(predicate, .T): unable to infer type `.T`; pass the type `.T` or bounds