An extension of an interior-point method for entropy minimization

Entropy optimization is used in signal compression, coding, estimation, and resource scheduling, among other applications. The paper presents a novel algorithm for entropy optimization. The algorithm is motivated by the efficient interior-point methods developed in Linear Programming. The algorithm uses a Generalized Affine Scaling Transformation that is an extension of the Affine Scaling Transformation utilized in interior-point methods. I show that for some entropy functions the proposed algorithm has superior convergence properties when compared to comparable the interior-point methods. The proposed algorithm is also shown to be related to, and a more general case of, the recently developed FOCUSS algorithm.