Trace norm regularization for multi-task learning with scarce data
This work addresses the challenge of learning from limited data in multi-task settings, offering theoretical guarantees for improved performance in data-scarce scenarios.
The paper tackled the problem of multi-task learning with scarce data by providing the first estimation error bound for trace norm regularization when samples per task are few, showing advantages that extend to meta-learning and are confirmed on synthetic datasets.
Multi-task learning leverages structural similarities between multiple tasks to learn despite very few samples. Motivated by the recent success of neural networks applied to data-scarce tasks, we consider a linear low-dimensional shared representation model. Despite an extensive literature, existing theoretical results either guarantee weak estimation rates or require a large number of samples per task. This work provides the first estimation error bound for the trace norm regularized estimator when the number of samples per task is small. The advantages of trace norm regularization for learning data-scarce tasks extend to meta-learning and are confirmed empirically on synthetic datasets.