Gradient-Based Transfer Learning

Gustaf Tegnér, Alfredo Reichlin, Hang Yin, Hedvig Kjellstrom, Mårten Björkman, Danica KragicICLR 2023 · OpenReview

Abstract

We formulate transfer learning as a meta-learning problem by extending upon the current meta-learning paradigm in that support and query data are drawn from different, but related distributions of tasks. Inspired by the success of Gradient-Based Meta-Learning (GBML), we propose to expand it to the transfer learning setting by constructing a general encoder-decoder architecture that learns a map between functionals of different domains. This is achieved by leveraging on the idea that the task-adapted parameters of a meta-learner can serve as an informative representation of the task itself. We demonstrate the proposed method on regression, prediction of dynamical systems and meta-imitation learning problems.
BibTeX
@inproceedings{tegner2023gradient,
  author = {Gustaf Tegnér and Alfredo Reichlin and Hang Yin and Hedvig Kjellstrom and Mårten Björkman and Danica Kragic},
  title = {{Gradient-Based Transfer Learning}},
  booktitle = {International Conference on Learning Representations},
  year = {2023},
  url = {https://openreview.net/forum?id=hChYEyebNm1},
}

Loading related content...

Figure