Philipp Becker

TL;DR: This work proposes a new deep approach to Kalman filtering which can be learned directly in an end-to-end manner using backpropagation without additional approximations and uses a high-dimensional factorized latent state representation for which the Kalman updates simplify to scalar operations and thus avoids hard to backpropagate, computationally heavy and potentially unstable matrix inversions.

TL;DR: In this article, a variational upper bound to the I-projection objective is proposed to decompose the original objective into single objectives for each mixture component as well as for the coefficients.

TL;DR: Two architectures are presented, one for forward model learning and one for inverse model learning, which significantly outperform existing model learning frameworks as well as analytical models in terms of prediction performance on a variety of real robot dynamics models.

TL;DR: The Recurrent Kalman Network (RKN) as mentioned in this paper uses a high-dimensional factorized latent state representation for which the Kalman updates simplify to scalar operations and thus avoids hard to backpropagate, computationally heavy and potentially unstable matrix inversions.

TL;DR: An alternative approach building on well-understood components for modeling aleatoric and epistemic uncertainty, dubbed Variational Recurrent Kalman Network (VRKN) is proposed, which uses Kalman updates for exact smoothing inference in a latent space and Monte Carlo Dropout to model epistem uncertainty.