DynaNet's differentiable Kalman filter on learned latent features improves visual odometry and motion prediction over LSTM baselines, with a Dirichlet-resampled transition matrix for stability.
Sequential monte carlo methods for dynamic systems,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2019 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
DynaNet: Neural Kalman Dynamical Model for Motion Estimation and Prediction
DynaNet's differentiable Kalman filter on learned latent features improves visual odometry and motion prediction over LSTM baselines, with a Dirichlet-resampled transition matrix for stability.