THESIS
2022
1 online resource (xii, 60 pages) : illustrations (some color)
Abstract
Active learning (AL) aims to improve model performance within a fixed labeling budget
by choosing the most informative data points to label. Existing AL focuses on the
single-domain setting, where all data come from the same domain (e.g., the same dataset).
However, many real-world tasks often involve multiple domains. For example, in visual
recognition, it is often desirable to train an image classifier that works across different environments
(e.g., different backgrounds), where images from each environment constitute
one domain. Such a multi-domain AL setting is challenging for prior methods because they
(1) ignore the similarity among different domains when assigning labeling budget and (2)
fail to handle distribution shift of data across different domains. Therefore, in this paper,...[
Read more ]
Active learning (AL) aims to improve model performance within a fixed labeling budget
by choosing the most informative data points to label. Existing AL focuses on the
single-domain setting, where all data come from the same domain (e.g., the same dataset).
However, many real-world tasks often involve multiple domains. For example, in visual
recognition, it is often desirable to train an image classifier that works across different environments
(e.g., different backgrounds), where images from each environment constitute
one domain. Such a multi-domain AL setting is challenging for prior methods because they
(1) ignore the similarity among different domains when assigning labeling budget and (2)
fail to handle distribution shift of data across different domains. Therefore, in this paper, we
propose the first general method for multi-domain AL. Our approach involves estimation
of domain similarity and reduction of distribution shift in the feature space via optimizing
an upper error bound that we develop for average errors of all domains. Our theoretical
analysis shows that our method achieves a better error bound compared to current AL
methods. Our empirical results demonstrate that our approach significantly outperforms
the state-of-the-art AL methods on both synthetic and real-world multi-domain datasets.
Post a Comment