Multi-Source Domain Adaptation with Mixture of Experts
read the original abstract
We propose a mixture-of-experts approach for unsupervised domain adaptation from multiple sources. The key idea is to explicitly capture the relationship between a target example and different source domains. This relationship, expressed by a point-to-set metric, determines how to combine predictors trained on various domains. The metric is learned in an unsupervised fashion using meta-training. Experimental results on sentiment analysis and part-of-speech tagging demonstrate that our approach consistently outperforms multiple baselines and can robustly handle negative transfer.
This paper has not been read by Pith yet.
Forward citations
Cited by 2 Pith papers
-
Closing the Domain Gap in Biomedical Imaging by In-Context Control Samples
Meta-learning with in-context control samples closes the domain gap for mechanism-of-action classification, raising accuracy on new batches from 0.862 to 0.935 on the JUMP-CP dataset.
-
TabICL: A Tabular Foundation Model for In-Context Learning on Large Data
TabICL scales in-context learning to large tabular data via column-then-row attention for row embeddings followed by a transformer, matching TabPFNv2 speed and performance while outperforming it and CatBoost on datase...
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.