Research map / Architectures and efficiency / Attention and transformers
End-To-End Multi-Task Learning With Attention
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 |
|---|---|---|---|
| Use task-specific soft attention masks over a shared feature pool | SegNet multi-task learning · NYUv2 validation | mIoU: 16.48 -> 17.72 | Helped |
| Use task-specific soft attention masks over a shared feature pool | SegNet multi-task learning · CityScapes 7-class validation | mIoU: 51.91 -> 53.86 | Helped |
| Use task-specific soft attention masks over a shared feature pool | SegNet multi-task learning · CityScapes 7-class validation | Rel Err: 25.36 -> 33.63 | Hurt |
| Use task-specific soft attention masks over a shared feature pool | SegNet multi-task learning · NYUv2 validation | Surface normal mean angle distance: 31.68 -> 31.25 | Helped |
| Use DWA task weighting instead of equal weights | MTAN · NYUv2 validation | mIoU: 17.72 -> 17.15 | Hurt |
| Use uncertainty weighting instead of equal weights | MTAN · NYUv2 validation | Pixel accuracy: 55.32 -> 55.61 | Helped |
| Use uncertainty weighting instead of equal weights | MTAN · NYUv2 validation | Surface normal mean angle distance: 31.44 -> 31.25 | Helped |
| Use DWA task weighting instead of equal weights | MTAN · NYUv2 validation | Rel Err: 0.2577 -> 0.2569 | Helped |
| Use MTAN rather than Parallel SVD | Visual Decathlon Challenge · Visual Decathlon Challenge online test | Score: 3398 -> 2941 | Hurt |
Closest papers by meaning
- Bridging Multi-Task Learning and Meta-Learning: Towards Efficient Training and Effective Adaptation (ICML 2021 · similarity 0.88)
- Integrated perception with recurrent multi-task neural networks (NeurIPS 2016 · similarity 0.88)
- Latent Multi-Task Architecture Learning (AAAI 2019 · similarity 0.88)
- Multi-Task Recurrent Modular Networks (AAAI 2021 · similarity 0.87)
- Cross-Stitch Networks for Multi-Task Learning (CVPR 2016 · similarity 0.87)
- Learn to Pay Attention (ICLR 2018 · similarity 0.87)
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.