\(\sin (\omega x)\) can approximate almost every finite set of samples (Q745856)

``Suppose \(x_1,\ldots, x_n\) are distinct with \(0\leq x_i\leq 1\) for \(i=1,\ldots,n\) and that \(-1<y_i<1\) for \(i=1,\ldots,n\). We are interested in the question of when a function of the form \[ y=\sin(\omega x) \] can approximate the set of points \((x_1,y_1),\ldots,(x_n,y_n)\) arbitrarily closely for a suitable choice of \(\omega\). That is, we address the question of whether for every \(\varepsilon >0\) there exists \(\omega\) such that \[ |\sin(\omega x_i)-y_i|<\varepsilon \text{ for } i=1,\ldots,n. \] In Sect. 2, we obtain a necessary and sufficient condition on the \(x_1,\ldots,x_n\) such that for \textit{any} \(y_1,\ldots,y_n\) with \(-1<y_i<1\), the points \((x_1,y_1),\ldots,(x_n,y_n)\) can be approximated. We call such \(x_1,\ldots,x_n\) \textit{approximable}, and we show that \(x_1,\ldots,x_n\) are \textit{approximable} if and only if the set \(\{x_1,\ldots,x_n\}\) is linearly independent over the rationals. This characterization follows straightforwardly from results in Diophantine analysis. In Sect. 3, we provide a different sufficient condition on the \(x_i\) (which we call mutually exhaustive) that guarantees approximability for any \(y_i\). Although this condition is not necessary, it gives a constructive way to select a suitable \(\omega\) for a desired level of approximation in terms of the binary expansions of the \(x_i\). Moreover, we show that if the \(x_i\) are chosen randomly (i.i.d. uniform), then the \(x_i\) will be mutually exhaustive almost surely.''