Directed KKL inequality

Let f ⁣:{0,1}n{0,1}f\colon\{0,1\}^n\to\{0,1\} be a Boolean function. Define its distance to monotonicity by

ε(f):=ming monotonePrx{0,1}n[f(x)g(x)].\varepsilon(f):=\min_{g\ \mathrm{monotone}}\Pr_{x\sim\{0,1\}^n}[f(x)\neq g(x)].

For i[n]i\in[n], define the negative influence of coordinate ii on ff by

Infi[f]:=#{x:f(x)>f(xi) and xxi}12n1.\operatorname{Inf}^{-}_i[f]:=\#\{x:f(x)>f(x^{\oplus i})\text{ and }x\preccurlyeq x^{\oplus i}\}\cdot\frac{1}{2^{n-1}}.

Directed KKL inequality. Given a Boolean function f ⁣:{0,1}n{0,1}f\colon\{0,1\}^n\to\{0,1\}, there exists i[n]i\in[n] such that

Infi[f]Ω(ε(f)lognn).\operatorname{Inf}^{-}_i[f]\geq\Omega\left(\varepsilon(f)\cdot\frac{\log n}{n}\right).

This is the natural directed analogue of the Kahn–Kalai–Linial inequality, with distance to monotonicity replacing variance and negative influences replacing ordinary influences. The supplied material does not establish whether the inequality is proved or remains open.

Sources & referencesView supporting material

Primary source

Quentin Dubroff, Shivam Nadimpalli and Bhargav Narayanan, “A Counterexample to a Directed KKL Inequality”, arXiv:2210.02035 (2022).

Progress summary

Never refreshed

Nothing recorded yet. Refresh searches the literature and the public web for attempts on this problem, and writes the first summary here.

Solutions 0

No solutions have been posted yet.