Research map / Architectures and efficiency / RNNs and normalization
SNR: Sub-Network Routing for Flexible Parameter Sharing in Multi-Task Learning
Jiaqi Ma, Zhe Zhao, Jilin Chen, Ang Li, Lichan Hong, Ed H. Chi
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 shared-bottom with SNR-Trans sub-network routing | Shared-Bottom MTL · YouTube8M | Macro Average-MAP@10 | Improved |
| replace MMoE with SNR-Aver | MMoE · YouTube8M | Macro Average-MAP@10 | Improved |
| stack multiple gated MMoE layers (ML-MMoE) | MMoE · YouTube8M | Macro Average-MAP@10 | Hurt |
| increase model size | SNR vs baselines · YouTube8M | Macro Average-MAP@10 | Improved |
| stronger L0 regularization on coding variables (sparse) at equal training size (ablation) | SNR-Trans · YouTube8M | Macro Average-MAP@10 | Hurt |
| stronger L0 regularization under limited serving size (ablation) | SNR-Trans · YouTube8M | serving model size at same accuracy: up to 11% smaller | Improved |
Closest papers by meaning
- Latent Multi-Task Architecture Learning (AAAI 2019 · similarity 0.86)
- MLR-SNet: Transferable LR Schedules for Heterogeneous Tasks (TPAMI 2022 · similarity 0.85)
- K for the Price of 1: Parameter-efficient Multi-task and Transfer Learning (ICLR 2019 · similarity 0.85)
- AlphaNet: Improved Training of Supernets with Alpha-Divergence (ICML 2021 · similarity 0.85)
- Inductive biases of multi-task learning and finetuning: multiple regimes of feature reuse (NeurIPS 2024 · similarity 0.83)
- Subnet-Aware Dynamic Supernet Training for Neural Architecture Search (CVPR 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.
