Research map / Time series / Causal and generative time series
Recurrent Kalman Networks: Factorized Inference in High-Dimensional Deep Feature Spaces
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use the RKN instead of an LSTM or GRU | Pendulum filtering · Pendulum | Log Likelihood: 6.019 -> 6.182; 6.051 -> 6.182 | Helped |
| use a full covariance matrix instead of the factorized covariance (ablation) | RKN · Pendulum | Log Likelihood: 6.182 -> 6.161; 6.248 -> 6.197 | No clear effect |
| use a full transition matrix instead of a sparse banded transition (ablation) | RKN · Pendulum | Log Likelihood: 6.182 -> 6.248 | No clear effect |
| use RKN instead of KVAE smoothing | Pendulum image imputation · Pendulum image imputation (informed) | Log Likelihood: -13.337 -> - 12.782 | Helped |
| use RKN instead of KVAE smoothing | Pendulum image imputation · Pendulum image imputation (uninformed) | Log Likelihood: -38.170 -> - 12.788 | Helped |
| use RKN instead of LSTM or GRU | Multiple Pendulum filtering · Multiple Pendulums | Log Likelihood: 7.5224 -> 11.51; 7.541 -> 11.51 | Helped |
| use RKN instead of LSTM or GRU | Quad Link filtering · Quad Link (without noise) | Log Likelihood: 11.960 -> 14.534; 10.346 -> 14.534 | Helped |
| use RKN instead of LSTM or GRU | Quad Link filtering · Quad Link (with noise) | Log Likelihood: 5.21 -> 6.259; 4.696 -> 6.259 | Helped |
| use RKN instead of KVAE smoothing | Quad Link image imputation · Quad Link image imputation (informed) | Log Likelihood: -52.608 -> -44.470 | Helped |
| use RKN instead of DeepVO | KITTI visual odometry · KITTI (mean) | Rotation relative error (degrees): 6.12 -> 3.20 | Helped |
| use RKN instead of DeepVO | KITTI visual odometry · KITTI (mean) | Translation relative error (%): 5.96 -> 9.48 | Hurt |
| use RKN instead of LSTM or GRU | Pneumatic Brook Robot Arm prediction · Pneumatic Brook Robot Arm | Log Likelihood: 0.952 -> 4.930; 1.186 -> 4.930 | Helped |
Closest papers by meaning
- Backprop KF: Learning Discriminative Deterministic State Estimators (NeurIPS 2016 · similarity 0.88)
- Predictive State Recurrent Neural Networks (NeurIPS 2017 · similarity 0.88)
- Gated KalmaNet: A Fading Memory Layer through Test-time Ridge Regression (CVPR 2026 · similarity 0.88)
- Two-Layer Linear Auto-Regressive Models Estimate Latent States (ICML 2026 · similarity 0.87)
- Gated Inference Network: Inference and Learning State-Space Models (NeurIPS 2024 · similarity 0.87)
- Learning long range dependencies through time reversal symmetry breaking (NeurIPS 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.