jax_privacy.matrix_factorization.sensitivity

Library for computing sensitivity under multiple participations.

Functions

banded_lower_triangular_mask(n, num_bands)	Returns n x n lower-triangular {0, 1} matrix with b bands of 1s.
banded_symmetric_mask(n, num_bands)	Returns n x n symmetric {0, 1} matrix with 2b - 1 bands of 1s.
fixed_epoch_sensitivity(C, epochs)	Like fixed_epoch_sensitivity_for_X(), but takes the encoder C.
fixed_epoch_sensitivity_for_X(X, epochs)	Compute the sensitivity of X under (k,b)-fixed-epoch participation.
get_min_sep_sensitivity_upper_bound(C[, ...])	Like get_min_sep_sensitivity_upper_bound_for_X, but takes the encoder C.
get_min_sep_sensitivity_upper_bound_for_X(X)	Computes an upper bound on the min_sep sensitivity of X.
get_sensitivity_banded(C[, min_sep, ...])	Like get_sensitivity_banded_for_X(), but takes the encoder C.
get_sensitivity_banded_for_X(X[, min_sep, ...])	Computes the sensitivity of an X.
max_participation_for_linear_fn(x[, ...])	Returns max_u <x, u>, where u respects the given participation pattern.
minsep_true_max_participations(n, min_sep[, ...])	Returns the maximum number of participations for a min_sep pattern.
single_participation_sensitivity(C)	Returns the sensitivity of a matrix with a single participation.