Distributed Kalman Filter through Gaussian Perception Propagation
Authors: Danny Bickson, Ori Shental, Danny Dolev
Summary: Latest end result exhibits compute distributively and effectively the linear MMSE for the multiuser detection downside, utilizing the Gaussian BP algorithm. Within the present work, we prolong this development, and present that working this algorithm twice on the matching inputs, has a number of attention-grabbing interpretations. First, we present equivalence to computing one iteration of the Kalman filter. Second, we present that the Kalman filter is a particular case of the Gaussian info bottleneck algorithm, when the load parameter β=1. Third, we talk about the relation to the Affine-scaling interior-point methodology and present it’s a particular case of Kalman filter. In addition to of the theoretical curiosity of this linking estimation, compression/clustering and optimization, we permit a single distributed implementation of these algorithms, which is a extremely sensible and necessary activity in sensor and cell ad-hoc networks. Utility to quite a few downside domains contains collaborative sign processing and distributed allocation of assets in a communication community.
