Research map / Architectures and efficiency / RNNs and normalization
State-Regularized 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 |
|---|---|---|---|
| Add state regularization to LSTM | LSTM · Balanced parentheses test, large dataset; d∈[10,20], l≤100 | error rate: 0.334 -> 0.255 | Helped |
| Add state regularization with peephole connections | LSTM · Balanced parentheses test, large dataset; d∈[10,20], l≤100 | error rate: 0.334 -> 0.001 | Helped |
| Add peephole connections to state-regularized LSTM (ablation) | sr-LSTM · Balanced parentheses test, large dataset; d∈[10,20], l≤100 | error rate: 0.255 -> 0.001 | Helped |
| Add state regularization to LSTM | LSTM · Balanced parentheses test, large dataset; d=5, l≤200 | error rate: 0.002 -> 0.044 | Hurt |
| Add state regularization with peephole connections | LSTM · Balanced parentheses test, large dataset; d=10, l≤200 | error rate: 0.207 -> 0.004 | Helped |
| Add state regularization with peephole connections | LSTM · Balanced parentheses test, small dataset; d=20, l≤1000 | error rate: 0.519 -> 0.380 | Helped |
| Add state regularization to LSTM | LSTM · Wikipedia language modeling | perplexity (Test): 85.2 -> 82.7 | Helped |
| Add peephole connections to state-regularized LSTM (ablation) | sr-LSTM · Wikipedia language modeling | perplexity (Test): 82.7 -> 81.3 | Helped |
| Add peephole connections to LSTM without state regularization (ablation) | LSTM · Wikipedia language modeling | perplexity (Test): 85.2 -> 86.9 | Hurt |
| Use state regularization with peephole connections | LSTM · Wikipedia language modeling | perplexity (Test): 85.2 -> 81.3 | Helped |
| Train sr-LSTM-p with different centroid counts k (ablation) | sr-LSTM-p · Balanced parentheses, small dataset | error rate | No clear effect |
| Use peephole connections in state-regularized LSTM (ablation) | sr-LSTM · Balanced parentheses extrapolation | error rate | Helped |
Closest papers by meaning
- Learning and Generalization in RNNs (NeurIPS 2021 · similarity 0.90)
- Understanding and Controlling Memory in Recurrent Neural Networks (ICML 2019 · similarity 0.90)
- Decision-Guided Weighted Automata Extraction from Recurrent Neural Networks (AAAI 2021 · similarity 0.90)
- Fast-Slow Recurrent Neural Networks (NeurIPS 2017 · similarity 0.89)
- Recurrent neural networks: vanishing and exploding gradients are not the end of the story (NeurIPS 2024 · similarity 0.89)
- M$^2$RNN: Non-Linear RNNs with Matrix-Valued States for Scalable Language Modeling (NeurIPS 2026 · similarity 0.88)
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.