Research map / Architectures and efficiency / RNNs and normalization
Can recurrent neural networks warp time?
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 |
|---|---|---|---|
| chrono initialization of LSTM gate biases | LSTM · Adding task (T=750) | convergence speed: converges 7 times faster | Helped |
| chrono initialization of LSTM gate biases | LSTM · Copy task (T=500, 2000, variable) | cross-entropy loss | Helped |
| chrono initialization of LSTM gate biases | LSTM · Variable copy task T=500 | cross-entropy loss | No clear effect |
| chrono initialization of LSTM gate biases | LSTM · Permuted MNIST | best validation accuracy: 96.3% vs 95.4% | Helped |
| chrono initialization of LSTM gate biases | LSTM · Standard MNIST | validation error | No clear effect |
| chrono initialization of LSTM gate biases | LSTM · Text8 character prediction | test bits per character: 1.37 vs 1.38 | Helped |
| gated recurrent units (learnable gates) vs plain RNN and leaky RNN on warped/padded sequences (ablation) | RNN architectures · Pure warping and padding tasks | loss after 3 epochs | Helped |
| learnable constant leaky gate vs ungated RNN (ablation) | RNN · Uniform time warping | loss | Helped |
| learnable constant leaky gate vs gated RNN (per-unit input-dependent gate) (ablation) | Leaky RNN · Variable time warping | loss | Hurt |
| chrono initialization of gate biases in RHN | Recurrent highway network · Penn Treebank word-level | test perplexity | No clear effect |
Closest papers by meaning
- Towards Non-Saturating Recurrent Units for Modelling Long-Term Dependencies (AAAI 2019 · similarity 0.89)
- Improving the Gating Mechanism of Recurrent Neural Networks (ICML 2020 · similarity 0.89)
- Fast Saturating Gate for Learning Long Time Scales with Recurrent Neural Networks (AAAI 2023 · similarity 0.88)
- Recurrent neural networks: vanishing and exploding gradients are not the end of the story (NeurIPS 2024 · similarity 0.88)
- Emergent mechanisms for long timescales depend on training curriculum and affect performance in memory tasks (ICLR 2024 · similarity 0.86)
- On the Importance of Gating: Memorization vs. In-Context Learning in State Space Models (NeurIPS 2026 · 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.