Research map / Architectures and efficiency / RNNs and normalization
AdaNet: Adaptive Structural Learning of Artificial Neural Networks
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 |
|---|---|---|---|
| Adaptively grow network structure while optimizing its weights | AdaNet · CIFAR-10 binary label pairs | accuracy: AdaNet exceeds the strongest baseline on each pair | Helped |
| Adaptively learn a one-layer structure rather than select a fixed NN architecture | AdaNet · Criteo Click Rate Prediction | accuracy: 0.7811 -> 0.7846 | Helped |
| Use last-hidden-layer output standard deviations instead of the Rademacher-complexity upper bound (ablation) | AdaNet · CIFAR-10 deer-truck | accuracy: 0.9372 ± 0.0082 -> 0.9309 ± 0.0069 | No clear effect |
| Add complexity-weighted L1 regularization during candidate-subnetwork generation (ablation) | AdaNet · CIFAR-10 deer-truck | accuracy: 0.9372 ± 0.0082 -> 0.9336 ± 0.0075 | No clear effect |
| Restrict each new subnetwork's connections to the immediately preceding subnetwork (ablation) | AdaNet · CIFAR-10 deer-truck | accuracy: 0.9372 ± 0.0082 -> 0.9321 ± 0.0065 | No clear effect |
| Apply dropout to connections to previously added subnetworks (ablation) | AdaNet · CIFAR-10 deer-truck | accuracy: 0.9372 ± 0.0082 -> 0.9376 ± 0.0080 | No clear effect |
Closest papers by meaning
- Adaptive Neural Networks for Efficient Inference (ICML 2017 · similarity 0.86)
- The Dynamics of Learning: A Random Matrix Approach (ICML 2018 · similarity 0.85)
- Constructing Deep Neural Networks by Bayesian Network Structure Learning (NeurIPS 2018 · similarity 0.84)
- On the Interplay between Graph Structure and Learning Algorithms in Graph Neural Networks (ICML 2025 · similarity 0.83)
- ASGO: Adaptive Structured Gradient Optimization (NeurIPS 2025 · similarity 0.83)
- From Kernels to Features: A Multi-Scale Adaptive Theory of Feature Learning (ICML 2025 · similarity 0.83)
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.