Parity Function and Imbalance Lemma #
The parity function and its interaction with Boolean functions and sensitivity.
Main results #
parity_flipBit— flipping a bit negates the paritymoebius_parity_sum— the Möbius coefficient relates to parity-weighted sumfullDegree_imbalance— if f has full degree n, one "parity-sign class" has more than half the vertices
The parity-signed function: g(x) = f_pm(x) · parity(x).
Equations
- f.paritySigned x = f.pmOne x * Sensitivity.parity x
Instances For
If paritySigned is the same at x and flipBit x i, then f is sensitive there.
Proof: parity flips sign under flipBit. If f_pm · parity stays the same, f_pm must also flip, so f changes value.
Key identity: the parity-weighted sum relates to the top Möbius coefficient. ∑x f_pm(x) · parity(x) = 2 · (-1)^n · c{univ}(f).
This follows from expanding the Möbius coefficients and using the bijection between Boolean vectors and subsets.
If f has full degree (moebius at univ ≠ 0), the parity-weighted sum is nonzero.
Imbalance: if the parity-weighted sum is nonzero, then one of the two "parity-sign classes" has more than 2^{n-1} elements.