Switch functions

Richard R. Hall, Eli Hawkins, Bernard S. Kay

Published 2017-10-08Version 1

We define a switch function to be a function from an interval to $\{1,-1\}$ with a finite number of sign changes. (Special cases are the Walsh functions.) We prove that, given $n$ real-valued functions, $f_1, \dots, f_n$, in $L^1[0,1]$, there exists a single switch function, $\sigma$, with at most $n$ sign changes that is simultaneously orthogonal to all of them in the sense that $\int_0^1 \sigma(t)f_i(t)dt=0$, for all $i = 1, \dots , n$. Moreover, we prove that, for each $\lambda \in [0,1]$, there exists a unique switch function, $\sigma$, with $n$ switches such that $\int_0^1 \sigma(t) p(t) dt = \lambda \int_0^1 p(t)dt$ for every real polynomial $p$ of degree at most $n-1$. We also prove the same statement holds for every real even polynomial of degree at most $2n-2$. Furthermore, for each of these latter results, we write down, in terms of $\lambda$ and $n$, a degree $n$ polynomial whose roots are the switch points of $\sigma$; we are thereby able to compute these switch functions.