Research map / Optimization / Optimizers
Efficient Orthogonal Parametrisation of Recurrent Neural Networks Using Householder Reflections
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 the Householder-parameterized orthogonal transition matrix instead of the LSTM | RNN on Penn Tree Bank · Penn Tree Bank 1-step-ahead test | test cost (bpc): 1.68 vs 1.68 | No clear effect |
| Use the Householder-parameterized orthogonal transition matrix instead of the LSTM | RNN on Penn Tree Bank · Penn Tree Bank 5-step-ahead test | test cost (bpc): 3.85 vs 3.8 | Hurt |
| Use leaky_ReLU instead of OPLU (ablation) | oRNN on the copy task · Copying task | performance | Hurt |
| Replace leaky_ReLU with the real representation of modReLU (ablation) | oRNN on the copy task · Copying task | performance | No clear effect |
Closest papers by meaning
- On orthogonality and learning recurrent networks with long term dependencies (ICML 2017 · similarity 0.90)
- Cheap Orthogonal Constraints in Neural Networks: A Simple Parametrization of the Orthogonal and Unitary Group (ICML 2019 · similarity 0.89)
- Learning Unitary Operators with Help From u(n) (AAAI 2017 · similarity 0.89)
- Coordinate Descent on the Orthogonal Group for Recurrent Neural Network Training (AAAI 2022 · similarity 0.89)
- Improved memory in recurrent neural networks with sequential non-normal dynamics (ICLR 2020 · similarity 0.89)
- Recurrent neural networks: vanishing and exploding gradients are not the end of the story (NeurIPS 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.