Research map / Architectures and efficiency / RNNs and normalization
Stateful ODE-Nets using Basis Function Expansions
Alejandro Francisco Queiruga, N. Benjamin Erichson, Liam Hodgkinson, Michael W. Mahoney
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| add continuous-in-depth batch normalization (ablation) | Stateful ODE-Net · MNIST | average accuracy: 99.1% -> 99.4% | Improved |
| add continuous-in-depth batch normalization (ablation) | Stateful ODE-Net · CIFAR-10 shallow | average accuracy: 88.4% -> 90.4% | Improved |
| add continuous-in-depth batch normalization (ablation) | Stateful ODE-Net · CIFAR-10 deep | average accuracy: 10.0% -> 92.4% | Improved |
| compress by reducing basis functions and timesteps 8 to 4 | Stateful ODE-Net · MNIST | average accuracy: 99.4% -> 99.2% | No clear effect |
| compress further to 7K parameters | Stateful ODE-Net · MNIST | average accuracy: 99.4% -> 96.9% | Hurt |
| compress basis functions 8 to 4 | Stateful ODE-Net · CIFAR-10 shallow | average accuracy: 90.4% -> 89.9% | No clear effect |
| compress model trained with refinement training (c1) | Stateful ODE-Net · CIFAR-10 deep | average accuracy: 92.4% -> 91.8% | No clear effect |
| compress model trained without refinement training (c2) | Stateful ODE-Net · CIFAR-10 deep | average accuracy: 94.1% -> 60.5% | Hurt |
| train with refinement training and piecewise linear basis (c1 vs c2) | Stateful ODE-Net · CIFAR-10 deep | average accuracy: 94.1% -> 92.4% | Hurt |
| refinement training (c1) on CIFAR-100 | Stateful ODE-Net · CIFAR-100 | average accuracy: 79.1% -> 76.9% | Hurt |
| compress c2 on CIFAR-100 | Stateful ODE-Net · CIFAR-100 | average accuracy: 79.1% -> 48.9% | Hurt |
| compress c1 on CIFAR-100 | Stateful ODE-Net · CIFAR-100 | average accuracy: 76.9% -> 75.6% | No clear effect |
Closest papers by meaning
- Neural Ordinary Differential Equations (NeurIPS 2018 · similarity 0.89)
- Neural Differential Equations for Learning to Program Neural Nets Through Continuous Learning Rules (NeurIPS 2022 · similarity 0.88)
- Do Residual Neural Networks discretize Neural Ordinary Differential Equations? (NeurIPS 2022 · similarity 0.88)
- ANODEV2: A Coupled Neural ODE Framework (NeurIPS 2019 · similarity 0.88)
- Improving Neural ODE Training with Temporal Adaptive Batch Normalization (NeurIPS 2024 · similarity 0.87)
- Implicit regularization of deep residual networks towards neural ODEs (ICLR 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.
