I have attached a PDF which more elegantly describes this (hopefully) simple problem. In the course of developing a new theory as part of my graduate research I have encountered an expression U* (U^{T} * S * U)^{-1} *U^{T} where U contains random orthonormal vectors and S is positive-definite. Through brute force I have demonstrated that the product is invariant to the particular vectors one chooses to populate U, but I cannot prove why. Does anyone have any ideas?

Similar questions and discussions