Research map / Learning theory and algorithms / Prediction and decision losses
Learning Symbolic Model-Agnostic Loss Functions via Meta-Learning
Christian Raymond, Qi Chen, Bing Xue, Mengjie Zhang
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 |
|---|---|---|---|
| replace cross-entropy with EvoMAL loss | ResNet-18 · CIFAR-10 | error rate: 0.0658 -> 0.0528 | Improved |
| replace cross-entropy with EvoMAL loss | LeNet-5 · MNIST | error rate: 0.0203 -> 0.0100 | Improved |
| replace squared error with EvoMAL loss | MLP · California Housing | test MSE: 0.2326 -> 0.1723 | Improved |
| replace squared error with EvoMAL loss | MLP · Boston Housing | test MSE: 0.1304 -> 0.1314 | No clear effect |
| replace cross-entropy with EvoMAL loss | PyramidNet · CIFAR-100 | error rate: 0.2527 -> 0.2664 | Hurt |
| add gradient-based local search (unrolled differentiation) to genetic programming (ablation) | EvoMAL · CIFAR-10 ResNet-18 | error rate: 0.0818 -> 0.0528 | Improved |
| add gradient-based local search (unrolled differentiation) to genetic programming (ablation) | EvoMAL · CIFAR-100 WideResNet | error rate: 0.3382 -> 0.2276 | Improved |
| transfer EvoMAL loss learned on CIFAR-10 to CIFAR-100 | EvoMAL · CIFAR-100 AlexNet (transfer) | error rate: 0.5262 -> 0.5324 | Hurt |
| transfer EvoMAL loss learned on CIFAR-10 to CIFAR-100 | EvoMAL · CIFAR-100 ResNet-18 (transfer) | error rate: 0.2474 -> 0.2326 | Improved |
| use parametric neural-network loss (ML3) instead of symbolic loss for transfer | ML3 · CIFAR-100 ResNet-18 (transfer) | error rate: 0.2474 -> 0.6000 | Hurt |
| tune base learning rate for cross-entropy baseline and EvoMAL (ablation) | EvoMAL · CIFAR-10 AllCNN-C | error rate | Mixed |
Closest papers by meaning
- Loss Function Learning for Domain Generalization by Implicit Gradient (ICML 2022 · similarity 0.90)
- MetaSymNet: A Tree-like Symbol Network with Adaptive Architecture and Activation Functions (AAAI 2025 · similarity 0.86)
- Loss Functions and Operators Generated by f-Divergences (ICML 2025 · similarity 0.85)
- Symbolic regression via MDLformer-guided search: from minimizing prediction error to minimizing description length (ICLR 2025 · similarity 0.85)
- Transformers Implement Functional Gradient Descent to Learn Non-Linear Functions In Context (ICML 2024 · similarity 0.84)
- Automated Loss function Search for Class-imbalanced Node Classification (ICML 2024 · similarity 0.84)
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.
