Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Neural Complexity Measures
Yoonho Lee, Juho Lee, Sung Ju Hwang, Eunho Yang, Seungjin Choi
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 |
|---|---|---|---|
| Regularize task learning with Neural Complexity | image classification learners · MNIST | test accuracy | Improved |
| Regularize task learning with Neural Complexity | image classification learners · FMNIST | test accuracy | Improved |
| Regularize task learning with Neural Complexity | image classification learners · KMNIST | test accuracy | Improved |
| Regularize task learning with Neural Complexity | image classification learners · SVHN | test accuracy | Hurt |
| Regularize task learning with Neural Complexity | image classification learners · CIFAR-10 | test accuracy | Improved |
| Regularize the learner with Neural Complexity rather than MetaReg | sinusoid regression · Sinusoid regression | test loss | Improved |
| Use Huber loss instead of MSE to train Neural Complexity (ablation) | Neural Complexity on Omniglot · Omniglot 10-way 1-shot | accuracy | Improved |
| Add a bias term to the classification interaction layer (ablation) | Neural Complexity classification architecture · Omniglot 10-way 1-shot | accuracy | Improved |
| Include training loss in the classification interaction layer (ablation) | Neural Complexity classification architecture · Omniglot 10-way 1-shot | accuracy | Improved |
| Use the bilinear layer in the classification interaction (ablation) | Neural Complexity classification architecture · Omniglot 10-way 1-shot | accuracy | Improved |
| Transfer Neural Complexity regularization to a task learner with 4 layers | sinusoid regression task learner · Sinusoid regression with changed learner hyperparameters | test loss | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MetaReg — MetaReg: Towards Domain Generalization using Meta-Regularization (NeurIPS 2018) · better on 5 of 5 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- The role of over-parametrization in generalization of neural networks (ICLR 2019 · similarity 0.88)
- Abstraction Mechanisms Predict Generalization in Deep Neural Networks (ICML 2020 · similarity 0.88)
- Topological Generalization Bounds for Discrete-Time Stochastic Optimization Algorithms (NeurIPS 2024 · similarity 0.86)
- Exploring the Complexity of Deep Neural Networks through Functional Equivalence (ICML 2024 · similarity 0.85)
- Do Neural Networks Need Gradient Descent to Generalize? A Theoretical Study (NeurIPS 2025 · similarity 0.85)
- A two-scale Complexity Measure for Deep Learning Models (NeurIPS 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.
