Inverse problems and derivatives of determinants (Q2276929)

A direct network problem has known conductances \(c_ i\) on each edge. Kirchhoff's law \(A^ Ty=f\) and Ohm's law \(y=CAx\) yield the currents y and potentials x. The key matrices are the node-edge incidence matrix A, the diagonal matrix of \(c_ i\), and especially \(A^ TCA.\) In an inverse problem the \(c_ i\) are unknown. They are to be determined from probes - measuring outputs x or y from specific inputs. These experiments yield information about \(R=A(A^ TCA)^{-1}A^ T\)- for our probes they yield the diagonal entries of R or CR. We study existence and uniqueness of the \(c_ i\), making essential use of the concave function \(\log \det A^ TCA\).