Research map / Architectures and efficiency / RNNs and normalization
Reviving and Improving Recurrent Back-Propagation
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 CG-RBP instead of original RBP | Recurrent back-propagation · Continuous Hopfield Network, truncate step 10 | success rate: 1% -> 100% | Helped |
| Use Neumann-RBP instead of original RBP | Recurrent back-propagation · Continuous Hopfield Network, truncate step 10 | success rate: 1% -> 100% | Helped |
| Use CG-RBP instead of original RBP | Recurrent back-propagation · Continuous Hopfield Network, truncate step 20 | success rate: 4% -> 100% | Helped |
| Use Neumann-RBP instead of original RBP | Recurrent back-propagation · Continuous Hopfield Network, truncate step 20 | success rate: 4% -> 100% | Helped |
| Use CG-RBP or Neumann-RBP instead of original RBP | Recurrent back-propagation · Continuous Hopfield Network, truncate step 30 | success rate: 99% -> 100% | No clear effect |
| Use Neumann-RBP instead of BPTT | Citation-network recurrent model · Cora | test accuracy: 24.48 $\pm$ 6.6 -> \textbf{46.63} $\pm$ 8.3 | Helped |
| Use Neumann-RBP instead of ARTBP with uniform truncation | Citation-network recurrent model · Cora | test accuracy: 27.88 $\pm$ 3.2 -> \textbf{46.63} $\pm$ 8.3 | Helped |
| Use Neumann-RBP instead of ARTBP with truncated Poisson truncation | Citation-network recurrent model · Cora | test accuracy: 42.22 $\pm$ 7.1 -> \textbf{46.63} $\pm$ 8.3 | Helped |
| Use Neumann-RBP instead of TBPTT | Citation-network recurrent model · Cora | test accuracy: 46.55 $\pm$ 6.4 -> \textbf{46.63} $\pm$ 8.3 | No clear effect |
Closest papers by meaning
- Training Recurrent Neural Networks via Forward Propagation Through Time (ICML 2021 · similarity 0.90)
- Memory-Efficient Backpropagation Through Time (NeurIPS 2016 · similarity 0.89)
- 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 0.87)
- Efficient recurrent architectures through activity sparsity and sparse back-propagation through time (ICLR 2023 · similarity 0.87)
- Exploring the Promise and Limits of Real-Time Recurrent Learning (ICLR 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.