Research map / Architectures and efficiency / RNNs and normalization
Orthogonal Recurrent Neural Networks with Scaled Cayley Transform
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 scoRNN instead of LSTM | pixel-by-pixel classifier · Permuted MNIST | test accuracy: 0.920 -> 0.966 | Helped |
| use scoRNN instead of LSTM | pixel-by-pixel classifier · Unpermuted MNIST | test accuracy: 0.989 -> 0.985 | Hurt |
| use scoRNN instead of restricted-capacity uRNN | pixel-by-pixel classifier · Permuted MNIST | test accuracy: 0.953 -> 0.966 | Helped |
| use scoRNN instead of full-capacity uRNN | pixel-by-pixel classifier · Unpermuted MNIST | test accuracy: 0.974 -> 0.985 | Helped |
| use scoRNN instead of LSTM | speech prediction model · TIMIT evaluation set | MSE: 14.30 -> 8.50 | Helped |
| use scoRNN instead of full-capacity uRNN | speech prediction model · TIMIT evaluation set | MSE: 14.58 -> 8.50 | Helped |
| use scoRNN instead of full-capacity uRNN | speech reconstruction · TIMIT | STOI | Hurt |
| use scoRNN instead of full-capacity uRNN | speech reconstruction · TIMIT | PESQ | Hurt |
| use scoRNN instead of restricted-capacity uRNN | recurrent model · Unpermuted MNIST | minutes per epoch: 8.2 -> 5.3 | Helped |
| use scoRNN instead of full-capacity uRNN | recurrent model · Unpermuted MNIST | minutes per epoch: 10.8 -> 5.3 | Helped |
| increase scoRNN hidden units from 170 to 512 (ablation) | scoRNN · Permuted MNIST | test accuracy: 0.943 -> 0.966 | Helped |
| increase scoRNN hidden units from 170 to 512 (ablation) | scoRNN · Unpermuted MNIST | test accuracy: 0.973 -> 0.985 | Helped |
Closest papers by meaning
- Complex Unitary Recurrent Neural Networks Using Scaled Cayley Transform (AAAI 2019 · similarity 0.96)
- Eigenvalue Normalized Recurrent Neural Networks for Short Term Memory (AAAI 2020 · similarity 0.90)
- Improved memory in recurrent neural networks with sequential non-normal dynamics (ICLR 2020 · similarity 0.90)
- Recurrent neural networks: vanishing and exploding gradients are not the end of the story (NeurIPS 2024 · similarity 0.86)
- HADAMRNN: BINARY AND SPARSE TERNARY ORTHOGONAL RNNS (ICLR 2025 · similarity 0.86)
- Learning Useful Representations of Recurrent Neural Network Weight Matrices (ICML 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.