Generalized Vandermonde Matrices with Missing Row-Powers

Let [n] := \{0, 1, 2, \cdots, n -1 \}. Let x = (x_0, \cdots, x_{n-1}) with distinc x_js. Let R = \{0 = r_1, r_2, \cdots, r_{n-1} = r be a set of distinct nonnegative row-powers. Consider the n \times n Generalized Vandermonde matrix V(x ; R) := \left( x_j^{r_i} \right)_{i,j \in [n]}. When R = [n], this matrix becomes the ordinary Vandermonde matrix, V(x).

An equivalent description of R is the largest row-power r \in R and the set of missing rows from [r]: that is, the items that are in [r] but not in R. Let L_r = \{\ell_0, \ell_1, \cdots \} be this set. Define the punctured Generalized Vandermonde matrix V_{\perp}(x; L_r) := V(x; R). Let s_k(x) be the kth elementary symmetric polynomial.

Now we are ready to present some interesting results without proof, from the paper “Lower bound on Sequence Complexity” by Kolokotronis, Limniotis, and Kalouptsidis.

det V_{\perp}(x ; \{\ell \}) = det V(x) \ s_{n-\ell}(x).

det V_{\perp}(x ; \{\ell_0, \ell_1\}) = det V(x) \ det \left( s_{n-\ell_i+j}(x) \right)_{i,j \in [2]}.

det V_{\perp}(x ; L) = det V(x) \ det \left( s_{n-\ell_i+j}(x) \right)_{i,j \in [s]} \text{ where } L = \{ \ell_0, \ell_1, \cdots, \ell_{s-1}\}.

I will add some applications later on.