Sign patterns of inverse doubly nonnegative matrices and inverse completely positive matrices (Q2109065)

A matrix is called \textit{doubly nonnegative} (shortly DNN) if it is positive semidefinite and entrywise nonnegative. It is called \textit{completely positive} (shortly CP) if it has a factorization of the form \(BB^T\) with \(B\geq 0\). The author proves that all inverses of DNN and CP realizations of a connected graph have the same \(\left\{+,-,0\right\}\) sign pattern if and only if the graph is bipartite. This result for the DNN matrices is insipired by [\textit{S. Roy} and \textit{M. Xue}, Linear Algebra Appl. 610, 480--487 (2021; Zbl 1458.15060)]. Moreover, the authors characterize the sign patterns of inverse DNN matrices that determine the graph of the initial DNN matrix. The analogous question for CP matrices remains open.