Research map / Probabilistic and scientific modeling / Bayesian optimization
Hyperparameter Optimization through Neural Network Partitioning
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 |
|---|---|---|---|
| learn affine augmentations with partitioned networks | ResNet-8 · CIFAR10 | test accuracy: 84.2_{\pm 0.8} -> 86.1_{\pm0.4} | Helped |
| learn affine augmentations with partitioned networks | ResNet-8 · RotCIFAR10 | test accuracy: 79.5_{\pm0.6} -> 79.1_{\pm0.0} | Hurt |
| learn affine augmentations with partitioned networks | ResNet-50 · TinyImagenet | test accuracy: 44.2_{\pm0.5} -> 48.6_{\pm0.0} | Helped |
| learn affine augmentations with partitioned networks | ResNet-50 · RotTinyImagenet | test accuracy: 44.5_{\pm0.2} -> 43.9_{\pm0.3} | Hurt |
| learn hyperparameters using partitioned networks | federated MNIST models · MNIST 1.25k | validation accuracy: 95.4\%_{\pm0.1} -> 97.0\%_{\pm0.1} | Helped |
| learn hyperparameters using partitioned networks | federated rotated-image models · RotMNIST 1.25k | validation accuracy: 80.5\%_{\pm0.0} -> 85.7\%_{\pm0.9} | Helped |
| learn hyperparameters using partitioned networks | federated CIFAR10 models · CIFAR10 45k | validation accuracy: 79.2\%_{\pm0.7} -> 81.5\%_{\pm0.5} | Helped |
| learn hyperparameters using partitioned networks | federated rotated CIFAR10 models · RotCIFAR10 45k | validation accuracy: 53.9\%_{\pm1.1} -> 60.6\%_{\pm0.2} | Helped |
| use partitioned training instead of validation-set optimization with finetuning | ResNet-14 affine augmentation learning · CIFAR10 20k, chunks [0.5,0.5] | test accuracy: 81.3\%_{\pm0.09} -> 83.0\%_{\pm0.01} | Helped |
| use partitioned training instead of validation-set optimization | Wide ResNet-20 with first two stages tuned · CIFAR10 | test accuracy: 59.6\%_{\pm0.6} -> 87.3\%_{\pm0.8} | Helped |
| vary the parameter and data-chunk partition proportions (ablation) | Partitioned networks · MNIST affine augmentation learning | test accuracy | Mixed |
| replace random weight partitioning with node partitioning (ablation) | Partitioned networks · initial experiments | performance | No clear effect |
Closest papers by meaning
- Reshuffling Resampling Splits Can Improve Generalization of Hyperparameter Optimization (NeurIPS 2024 · similarity 0.89)
- On the Parameterization of Second-Order Optimization Effective towards the Infinite Width (ICLR 2024 · similarity 0.86)
- Magnitude Invariant Parametrizations Improve Hypernetwork Learning (ICLR 2024 · similarity 0.86)
- Principled Architecture-aware Scaling of Hyperparameters (ICLR 2024 · similarity 0.86)
- Sparse maximal update parameterization: A holistic approach to sparse training dynamics (NeurIPS 2024 · similarity 0.86)
- Generalization Bounds via Meta-Learned Model Representations: PAC-Bayes and Sample Compression Hypernetworks (ICML 2025 · 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.