Learning to select data for transfer learning with Bayesian Optimization
This work addresses data selection for transfer learning, which is an incremental improvement over existing domain similarity measures.
The paper tackled the problem of selecting suitable data for transfer learning by learning data selection measures using Bayesian Optimization, outperforming existing domain similarity measures significantly on sentiment analysis, part-of-speech tagging, and parsing tasks.
Domain similarity measures can be used to gauge adaptability and select suitable data for transfer learning, but existing approaches define ad hoc measures that are deemed suitable for respective tasks. Inspired by work on curriculum learning, we propose to \emph{learn} data selection measures using Bayesian Optimization and evaluate them across models, domains and tasks. Our learned measures outperform existing domain similarity measures significantly on three tasks: sentiment analysis, part-of-speech tagging, and parsing. We show the importance of complementing similarity with diversity, and that learned measures are -- to some degree -- transferable across models, domains, and even tasks.