Research map / Time series / Causal and generative time series
Sparse Gaussian Conditional Random Fields on Top of Recurrent Neural Networks
Xishun Wang, Minjie Zhang, Fenghui Ren
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 |
|---|---|---|---|
| add CoR (SGCRF on top of Bi-LSTM) to stacked Bi-LSTMs | stacked Bi-LSTMs · large synthetic | MAPE: 0.032 -> 0.026 | Improved |
| add CoR on top of Bi-LSTMs | stacked Bi-LSTMs · large synthetic | time usage: 5.43h -> 9.41h | Hurt |
| use stacked Bi-LSTMs instead of SGCRF | SGCRF · large synthetic | memory usage: 17.21G -> 3.53G | Improved |
| use stacked Bi-LSTMs instead of SGCRF | SGCRF · large synthetic | MAPE: 0.065 -> 0.032 | Improved |
| end-to-end training instead of alternative training (ablation) | CoR(LSTM) · NPower Forecasting Challenge 2016 round 1 | MAPE: (4.11 ± 0.04)% -> (4.05 ± 0.05)% | Improved |
| end-to-end training instead of alternative training (ablation) | CoR(RNN) · NPower Forecasting Challenge 2016 round 1 | MAPE: (4.25 ± 0.04)% -> (4.19 ± 0.06)% | Improved |
| end-to-end training instead of alternative training (ablation) | CoR(LSTM) · NPower Forecasting Challenge 2016 round 2 | MAPE: (3.94 ± 0.03)% -> (3.87 ± 0.03)% | Improved |
| feature engineering for SGCRF | SGCRF · NPower Forecasting Challenge 2016 round 1 | MAPE: 5.83% -> 4.91% | Improved |
| add CoR to stacked Bi-RNNs | stacked Bi-RNNs · NPower Forecasting Challenge 2016 round 1 | MAPE: (4.88 ± 0.21)% -> (4.19 ± 0.06)% | Improved |
| add CoR to stacked Bi-LSTMs | stacked Bi-LSTMs · NPower Forecasting Challenge 2016 round 2 | MAPE: (4.31 ± 0.14)% -> (3.87 ± 0.03)% | Improved |
| replace stacked Bi-RNNs with stacked Bi-LSTMs | stacked Bi-RNNs · NPower Forecasting Challenge 2016 round 1 | MAPE: (4.88 ± 0.21)% -> (4.43 ± 0.16)% | Improved |
| add feature engineering to gradient boosting | Gradient boosting · NPower Forecasting Challenge 2016 | MAPE: 4.90% | No clear effect |
Closest papers by meaning
- Sparse Deep Learning for Time Series Data: Theory and Applications (NeurIPS 2023 · similarity 0.86)
- Dynamical Isometry and a Mean Field Theory of RNNs: Gating Enables Signal Propagation in Recurrent Neural Networks (ICML 2018 · similarity 0.85)
- One-Shot Pruning of Recurrent Neural Networks by Jacobian Spectrum Evaluation (ICLR 2020 · similarity 0.85)
- Sparse Spiking Neural Network: Exploiting Heterogeneity in Timescales for Pruning Recurrent SNN (ICLR 2024 · similarity 0.83)
- STanHop: Sparse Tandem Hopfield Model for Memory-Enhanced Time Series Prediction (ICLR 2024 · similarity 0.83)
- Dynamic Sparse No Training: Training-Free Fine-tuning for Sparse LLMs (ICLR 2024 · similarity 0.83)
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.
