jax_privacy.matrix_factorization.banded

Class and instances for expressing and optimizing banded strategies.

Functions

minsep_sensitivity_squared(strategy, min_sep)	Returns the sensitivity of the ColumnNormalizedBanded strategy.
optimize(n, *, bands[, C, A, ...])	Optimize the strategy using a gradient-based method.
per_query_error(C[, A, scan_fn])	Computes expected per-query squared error of a strategy.

Classes

ColumnNormalizedBanded(params)

A column-normalized banded lower triangular n x n matrix.