Options
2018
Conference Paper
Titel
Time evolution for dynamic probabilistic tensors in hierarchical tucker decomposition form
Abstract
Many real world applications of target tracking and state estimation are non-linear filtering problems and can therefore not be solved by closed-form analytical solutions. In the recent past, tensor based approaches have become increasingly popular due to very effective decomposition algorithms, which allow the representation of discretized, high-dimensional data in compressed form. In this paper, a solution of the prediction step for a Bayesian filter is proposed, where the probability density function (pdf) is approximated by a tensor in Hierarchical Tucker Decomposition. It is shown, that the computation of the predicted pdf is about five times faster than the previously proposed Canonical Polyadic Decomposition format.