Improved maximum entropy analysis with an extended search space

Alexander Rothkopf Fakultät für Physik, Universität Bielefeld, D-33615 Bielefeld, Germany (Dated: October 31, 2011) Abstract We propose an improvement to the implementation of the well established Maximum Entropy Method by taking into account carefully the role of prior information. This involves a departure from Bryan’s concept of search space, a change that becomes crucial for the success of the inversion task in case of e.g. small datasets or unreliable prior information. To adequately approach problems where an exponentially damped Kernel is used, we provide an implementation using the C/C++ language that utilizes high precision arithmetic adjustable at runtime [1]. The LBFGS algorithm is included in the code in order to attack problems without the need to resort to any search space restriction. An analysis of mockas well as lattice QCD data is presented as illustration of the discussed issues.