Research map / Optimization / Optimizers
Amortized Proximal Optimization
Juhan Bae, Paul Vicol, Jeff Z. HaoChen, Roger Baker Grosse
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 |
|---|---|---|---|
| meta-learn structured preconditioner with APO instead of SGDm | SGDm · CIFAR-10 LeNet | test accuracy: 75.73 -> 77.42 | Improved |
| meta-learn structured preconditioner with APO instead of SGDm | SGDm · CIFAR-100 AlexNet | test accuracy: 43.95 -> 52.35 | Improved |
| APO-Precond instead of SGDm on batch-norm ResNets | SGDm · CIFAR-100 ResNet-32 | test accuracy: 77.47 -> 77.41 | No clear effect |
| APO-Precond instead of KFAC in 16-bit training | KFAC · CIFAR-10 ResNet-18 16-bit | test accuracy: 92.72 -> 94.79 | Improved |
| APO-Precond instead of KFAC in 16-bit training | KFAC · CIFAR-100 ResNet-18 16-bit | test accuracy: 73.12 -> 75.47 | Improved |
| APO-Precond instead of AdamW with fixed LR | AdamW · IWSLT14 De-En | BLEU: 34.60 -> 34.62 | No clear effect |
| APO-Precond instead of SGDm | SGDm · IWSLT14 De-En | BLEU: 31.43 -> 34.62 | Improved |
| tune learning rate with APO instead of best fixed LR (SGDm) | SGDm · CIFAR-10 ResNet-32 | test accuracy: 89.40 -> 92.75 | Improved |
| tune learning rate with APO instead of best fixed LR (SGDm) | SGDm · CIFAR-100 WRN 28-10 | test accuracy: 74.81 -> 79.33 | Improved |
| tune learning rate with APO instead of manual step decay (SGDm) | SGDm · CIFAR-100 WRN 28-10 | test accuracy: 81.01 -> 79.33 | Hurt |
| APO instead of manual decay for Adam | Adam · CIFAR-100 WRN 28-10 | test accuracy: 75.53 -> 76.33 | Improved |
| APO instead of manual decay for RMSprop | RMSprop · CIFAR-10 WRN 28-10 | test accuracy: 93.60 -> 94.22 | Improved |
Closest papers by meaning
- Adaptive Proximal Gradient Methods for Structured Neural Networks (NeurIPS 2021 · similarity 0.88)
- Learning Proximal Operators to Discover Multiple Optima (ICLR 2023 · similarity 0.88)
- What's in a Prior? Learned Proximal Networks for Inverse Problems (ICLR 2024 · similarity 0.86)
- Extensions to the Proximal Distance Method of Constrained Optimization (JMLR 2022 · similarity 0.86)
- Provable and Practical Online Learning Rate Adaptation with Hypergradient Descent (ICML 2025 · similarity 0.85)
- Structured Preconditioners in Adaptive Optimization: A Unified Analysis (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.
