Research map / Architectures and efficiency / RNNs and normalization
Approximating Real-Time Recurrent Learning with Random Kronecker Factors
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 KF-RTRL instead of UORO | RHN on the copy task · Copy task | maximum memorized string length: T=13 -> T=36 | Helped |
| Use KF-RTRL instead of UORO | RHN character-level language modeling · Penn TreeBank | performance | Helped |
| Use TBPTT instead of KF-RTRL | RHN character-level language modeling · Penn TreeBank | performance | Helped |
| Increase TBPTT truncation horizon from 5 to 25 steps | TBPTT on RHN character-level language modeling · Penn TreeBank | performance | No clear effect |
| Average n independent UORO gradient samples | UORO gradient estimation · Penn TreeBank variance analysis | gradient alignment as network size increases | Helped |
Closest papers by meaning
- Optimal Kronecker-Sum Approximation of Real Time Recurrent Learning (ICML 2019 · similarity 0.96)
- Exploring the Promise and Limits of Real-Time Recurrent Learning (ICLR 2024 · similarity 0.91)
- Practical Real Time Recurrent Learning with a Sparse Approximation (ICLR 2021 · similarity 0.91)
- Training Recurrent Neural Networks Online by Learning Explicit State Variables (ICLR 2020 · similarity 0.88)
- Recurrent neural networks: vanishing and exploding gradients are not the end of the story (NeurIPS 2024 · similarity 0.86)
- Second-order forward-mode optimization of recurrent neural networks for neuroscience (NeurIPS 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.