An orthogonal basis for functions over a slice of the Boolean hypercube

Abstract

We present a simple, explicit orthogonal basis of eigenvectors for the Johnson and Kneser graphs, based on Young’s orthogonal representation of the symmetric group. Our basis can also be viewed as an orthogonal basis for the vector space of all functions over a slice of the Boolean hypercube (a set of the form f(x1...,xn) ɛ {0, 1}n : ∑i xi= k}), whichBrefinesnthepeigenspacestoffthe Johnson1association scheme; our basis is orthogonal with respect to any exchangeable measure. More concretely, our basis is an orthogonal basis for all multilinear polynomials Rn → R which are annihilated by the differential operator ∑i ∂/∂xi. As an application of the last point of view, we show how to lift low-degree functions from a slice to the entire Boolean hypercube while maintaining properties such as expectation, variance and L2-norm. As an application of our basis, we streamline Wimmer’s proof of Friedgut’s theorem for the slice. Friedgut’s theorem, a fundamental result in the analysis of Boolean functions, states that a Boolean function on the Boolean hypercube with low total influence can be approximated by a Boolean junta (a function depending on a small number of coordinates). Wimmer generalized this result to slices of the Boolean hypercube, working mostly over the symmetric group, and utilizing properties of Young’s orthogonal representation. Using our basis, we show how the entire argument can be carried out directly on the slice.