Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Learning Scalable Deep Kernels with Recurrent Structure
Maruan Al-Shedivat, Andrew Gordon Wilson, Yunus Saatchi, Zhiting Hu, Eric P. Xing
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| replace LSTM output layer with a GP using the recurrent kernel | LSTM · Drives autoregression | RMSE: 0.14 -> 0.13 | Improved |
| add GP layer to LSTM | LSTM · Drives regression | RMSE: 0.29 -> 0.25 | Improved |
| add GP layer to LSTM | LSTM · Actuator regression | RMSE: 0.40 -> 0.36 | Improved |
| add GP layer to LSTM | LSTM · Actuator autoregression | RMSE: 0.19 -> 0.14 | Improved |
| add GP layer to LSTM | LSTM · Drives free simulation | RMSE: 0.40 -> 0.32 | Improved |
| add GP layer to LSTM | LSTM · Actuator free simulation | RMSE: 0.44 -> 0.43 | No clear effect |
| add GP layer to RNN | RNN · GEF power load | RMSE: 0.61 -> 0.24 | Improved |
| add GP layer to LSTM | LSTM · GEF power load | RMSE: 0.45 -> 0.17 | Improved |
| add GP layer to LSTM | LSTM · Car lanes | RMSE: 0.08 -> 0.06 | Improved |
| add GP layer to LSTM | LSTM · GEF wind power | RMSE: 0.77 -> 0.76 | No clear effect |
| add GP layer to NARX | NARX · GEF power load | RMSE: 0.54 -> 0.78 | Hurt |
| use RGP instead of GP-LSTM in free simulation | GP-LSTM · Drives free simulation | RMSE: 0.32 -> 0.249 | Improved |
Closest papers by meaning
- Kernel-Based Approaches for Sequence Modeling: Connections to Neural Methods (NeurIPS 2019 · similarity 0.89)
- Characterizing Deep Gaussian Processes via Nonlinear Recurrence Systems (AAAI 2021 · similarity 0.88)
- xLSTM: Extended Long Short-Term Memory (NeurIPS 2024 · similarity 0.86)
- Recurrent Memory for Online Interdomain Gaussian Processes (NeurIPS 2025 · similarity 0.86)
- Recurrent neural networks: vanishing and exploding gradients are not the end of the story (NeurIPS 2024 · similarity 0.85)
- Parallelizing non-linear sequential models over the sequence length (ICLR 2024 · similarity 0.85)
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.
