Superquadratic sensitivity separation: model #
This file fixes the model of Boolean functions f : (I → Bool) → Bool used throughout the
separation: flipping a finite set of input coordinates, the sensitivity of f at an input and
overall, and the block sensitivity of f at an input and overall. All maxima are Finset.sups
over the actual finite input space or the finite space of block families, so the definitions
agree literally with the standard ones. These measures are internal to the construction, which
needs arbitrary finite coordinate types; Bridge identifies them on Fin n with CSLib's
BitString.sensitivity and QueryComplexity.blockSensitivity.
Main definitions #
flip: negate the coordinates of an input lying in a given finite set; this isBitString.flipOnwith the arguments swapped and classical decidable equality.sensitivityAt: the number of coordinates whose flip changesf x.sensitivity: the maximum ofsensitivityAt f xover all inputsx.blockSensitivityAt: the maximum number of disjoint nonempty blocks whose flips changef x.blockSensitivity: the maximum ofblockSensitivityAt f xover all inputsx.
References #
Maximum size of a family of disjoint nonempty sensitive blocks at an input. The empty family is allowed and contributes zero. Invalid families contribute zero.