Adaptive and Robust Multi-Task Learning
This work addresses the challenge of learning from multiple datasets with potential outliers, which is important for applications in data analysis, but appears incremental as it builds on existing multi-task learning frameworks.
The paper tackles the multi-task learning problem by proposing adaptive methods that leverage similarities and handle differences among tasks, with proven statistical guarantees and robustness against outliers, demonstrating efficacy on synthetic and real datasets.
We study the multi-task learning problem that aims to simultaneously analyze multiple datasets collected from different sources and learn one model for each of them. We propose a family of adaptive methods that automatically utilize possible similarities among those tasks while carefully handling their differences. We derive sharp statistical guarantees for the methods and prove their robustness against outlier tasks. Numerical experiments on synthetic and real datasets demonstrate the efficacy of our new methods.