Research map / Architectures and efficiency / Pruning and efficient tuning
NestedNet: Learning Nested Sparse Structures in Deep 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 |
|---|---|---|---|
| learn nested weight masks instead of using random sparse masks (ablation) | NestedNet weight pruning · CIFAR-10, 2× compression | accuracy: 91.2% -> 93.4% | Helped |
| learn nested weight masks instead of using random sparse masks (ablation) | NestedNet weight pruning · CIFAR-10, 3× compression | accuracy: 85.1% -> 92.8% | Helped |
| use weight pruning rather than channel scheduling | NestedNet · CIFAR-10, 2× compression | accuracy: 92.9% -> 93.4% | Helped |
| use NestedNet channel scheduling rather than standalone channel-pruning methods | ResNet-56 · CIFAR-10, 2× compression | accuracy: 91.5% -> 92.9%; 91.8% -> 92.9% | Helped |
| train nested WRN-14-4 jointly with the full WRN-32-4 | NestedNet · CIFAR-100 knowledge distillation | accuracy: 73.8% -> 74.3% | Helped |
| train nested WRN-32-1 jointly with the full WRN-32-4 | NestedNet · CIFAR-100 knowledge distillation | accuracy: 67.5% -> 67.1% | Hurt |
| train nested WRN-14-1 jointly with the full WRN-32-4 | NestedNet · CIFAR-100 knowledge distillation | accuracy: 64.0% -> 64.3% | Helped |
| add the 20-class coarse hierarchy task to the nested full model | NestedNet WRN-14-8 · CIFAR-100 hierarchical classification, 100 classes | accuracy: 75.8% -> 77.3% | Helped |
| add the 20-class coarse hierarchy task to the nested full model | NestedNet WRN-14-4 · CIFAR-100 hierarchical classification, 20 classes | accuracy: 82.4% -> 83.9% | Helped |
| add the 20-class coarse hierarchy task to the nested full model | NestedNet WRN-32-4 · CIFAR-100 hierarchical classification, 100 classes | accuracy: 75.7% -> 76.6% | Helped |
| add the 20-class coarse hierarchy task to the nested full model | NestedNet WRN-32-2 · CIFAR-100 hierarchical classification, 20 classes | accuracy: 82.1% -> 83.7% | Helped |
| average outputs from nested levels | NestedNet-A · CIFAR-100 hierarchical classification, 100 classes | accuracy: 77.3% -> 78.3%; 76.6% -> 78.0% | Helped |
Closest papers by meaning
- Nested Learning: The Illusion of Deep Learning Architectures (NeurIPS 2025 · similarity 0.90)
- Deep Hierarchical Learning with Nested Subspace Networks (ICLR 2026 · similarity 0.87)
- Random Sparse Lifts: Construction, Analysis and Convergence of finite sparse networks (ICLR 2024 · similarity 0.85)
- MotherNet: Fast Training and Inference via Hyper-Network Transformers (ICLR 2025 · similarity 0.83)
- How Deep Networks Learn Sparse and Hierarchical Data: the Sparse Random Hierarchy Model (ICML 2024 · similarity 0.83)
- Inductive biases of multi-task learning and finetuning: multiple regimes of feature reuse (NeurIPS 2024 · 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.