Research map / Probabilistic and scientific modeling / Dynamical systems
Predictive State Recurrent Neural Networks
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 a rank-60 factorized PSRNN instead of an LSTM or GRU | PSRNN sequence prediction · Penn Tree Bank (PTB) | Bits Per Character (BPC) | Helped |
| Use a rank-60 factorized PSRNN instead of an LSTM or GRU | PSRNN sequence prediction · Penn Tree Bank (PTB) | one-step prediction accuracy (OSPA) | Helped |
| Add a second PSRNN layer (ablation) | PSRNN · Penn Tree Bank (PTB) | cross-entropy | Helped |
| Add a second PSRNN layer (ablation) | PSRNN · Penn Tree Bank (PTB) | prediction accuracy | Helped |
| Replace a PSRNN with a factorized PSRNN (ablation) | PSRNN · Penn Tree Bank (PTB) | cross-entropy | Hurt |
| Increase the factorization rank (ablation) | Factorized PSRNN · Penn Tree Bank (PTB) | cross-entropy | Helped |
| Initialize PSRNN weights randomly instead of with two-stage regression (ablation) | PSRNN · All evaluated datasets | MSE | Hurt |
| Replace full PSR normalization with two-norm normalization (ablation) | PSRNN initialized by two-stage regression · Model performance prior to refinement | prediction performance | No clear effect |
| Replace full PSR normalization with two-norm normalization (ablation) | PSRNN initialized by two-stage regression · Model performance after refinement | prediction performance | Helped |
| Use PSRNNs instead of alternative approaches, including LSTMs, GRUs, basic RNNs, and Kalman filters | Sequence prediction · Swimmer | MSE | Helped |
| Use PSRNNs instead of alternative approaches, including LSTMs, GRUs, basic RNNs, and Kalman filters | Sequence prediction · Mocap | MSE | Helped |
| Use PSRNNs instead of alternative approaches, including LSTMs, GRUs, basic RNNs, and Kalman filters | Sequence prediction · Handwriting | MSE | Helped |
Closest papers by meaning
- Predictive-State Decoders: Encoding the Future into Recurrent Networks (NeurIPS 2017 · similarity 0.93)
- Initialization matters: Orthogonal Predictive State Recurrent Neural Networks (ICLR 2018 · similarity 0.91)
- Learning dynamics in linear recurrent neural networks (ICML 2025 · similarity 0.88)
- Recurrent neural networks: vanishing and exploding gradients are not the end of the story (NeurIPS 2024 · similarity 0.86)
- Parallel-in-Time Training of Recurrent Neural Networks for Dynamical Systems Reconstruction (NeurIPS 2026 · similarity 0.86)
- Almost-Linear RNNs Yield Highly Interpretable Symbolic Codes in Dynamical Systems Reconstruction (NeurIPS 2024 · 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.