Research map / Architectures and efficiency / Neural architecture search
SplitNets: Designing Neural Architectures for Efficient Distributed Computing on Head-Mounted Systems
Check 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 |
|---|---|---|---|
| Jointly search the backbone, split position, and compression module | SplitNets · ImageNet | Overall Latency: 4.25ms -> 0.88ms | Helped |
| Jointly search the backbone, split position, and compression module | SplitNets · ModelNet40 | Top-1: 92.1 -> 93.8 | Helped |
| Use geometric-mean channel-based initialization instead of Kaiming Fan-In (ablation) | SplitNets supernet · ImageNet | Accuracy: +0.3% | Helped |
| Increase the base learning rate with conventional initialization (ablation) | SplitNets supernet · ImageNet | Training stability | Hurt |
| Use multiple splitting modules during supernet training rather than sandwich sampling with one-hot split variants (ablation) | Single-view SplitNets · ImageNet | Accuracy drop due to splitting modules | Helped |
| Use sandwich sampling for multi-view models, where only one fusion module can be sampled (ablation) | Multi-view SplitNets · ModelNet40 | Top-1 | No clear effect |
Closest papers by meaning
- HEP-NAS: Towards Efficient Few-shot Neural Architecture Search via Hierarchical Edge Partitioning (AAAI 2025 · similarity 0.82)
- FreeNet: Liberating Depth-Wise Separable Operations for Building Faster Mobile Vision Architectures (AAAI 2025 · similarity 0.82)
- HydraViT: Stacking Heads for a Scalable ViT (NeurIPS 2024 · similarity 0.82)
- Efficient Few-Shot Neural Architecture Search by Counting the Number of Nonlinear Functions (AAAI 2025 · similarity 0.81)
- LSNet: See Large, Focus Small (CVPR 2025 · similarity 0.81)
- An Empirical Study of Distributed Deep Learning Training on Edge (Student Abstract) (AAAI 2024 · similarity 0.81)
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.