Research map / Learning theory and algorithms / Matrix and tensor methods
Learning Sparse Task Relations in Multi-Task Learning
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 |
|---|---|---|---|
| replace dense task relations (MTL-TNR) with sparse l1 task covariance (SPATS) | multi-task model · Synthetic Data 2 | NMSE: 0.0401 -> 0.0280 | Helped |
| replace sparse inverse covariance (gLasso) with sparse covariance | multi-task model · Synthetic Data 1 | NMSE: 0.1013 -> 0.0646 | Helped |
| replace single-task learning with SPATS multi-task learning | ridge regression · School | NMSE: 1.1453 -> 0.9110 | Helped |
| replace STL with SPATS | ridge regression · Landmine | AUC: 0.6854 -> 0.7420 | Helped |
| replace dense (MTL-TNR/STNR) with sparse task relations when only 4 tasks | multi-task model · Sentiment | AUC: 0.8808 -> 0.8576 | Hurt |
| replace MTL-STNR with SPATS | multi-task model · School | NMSE: 0.9048 -> 0.9110 | No clear effect |
| replace MTL-STNR with SPATS | multi-task model · MHC-I | AUC: 0.6943 -> 0.7299 | Helped |
| replace MTL-STNR with SPATS | multi-task model · Landmine | AUC: 0.7496 -> 0.7420 | No clear effect |
| assume fixed identity task covariance (MTL-Omega) instead of STL on synthetic data | ridge regression · Synthetic Data 1 | NMSE: 0.1033 -> 0.1287 | Hurt |
Closest papers by meaning
- Asymmetric Multi-task Learning Based on Task Relatedness and Loss (ICML 2016 · similarity 0.87)
- Identifying and Mitigating Spurious Correlation in Multi-Task Learning (CVPR 2025 · similarity 0.87)
- Learning Multi-Task Sparse Representation Based on Fisher Information (AAAI 2024 · similarity 0.86)
- Joint-Task Regularization for Partially Labeled Multi-Task Learning (CVPR 2024 · similarity 0.86)
- Analysing Multi-Task Regression via Random Matrix Theory with Application to Time Series Forecasting (NeurIPS 2024 · similarity 0.84)
- Task Groupings Regularization: Data-Free Meta-Learning with Heterogeneous Pre-trained Models (ICML 2024 · similarity 0.84)
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.