Matrix Operations for the Simulation and Immediate Reverse-Engineering of Time Series Data

Abstract

We present a new method for constructing and decomposing square matrices. This method, based on the computed parameterisation of their implied determinants and minors, operates on the product of factors of a new form of matrix decomposition. This method may be employed to build new matrices with fixed determinant(s). We demonstrate that this new approach is fundamentally well-connected to the Cholesky decomposition if applied on symmetric matrices. We also demonstrate that it is related to the LU decomposition method via a diagonal matrix multiplier. Also through this new method a direct relation between Cholesky decomposition and LU factorisation is shown. This method, presented for the first time, is useful for (re)constructing matrices with a predefined determinant and simulating inverse problems. The inference method introduced here also is based on new matrix manipulation techniques that we have developed for the identification of systems from reproducible time series data.