two stages domain invariant representation learners solve the large co-variate shift in unsupervised domain adaptation with two dimensional data domains

Hisashi Oshima, Tsuyoshi Ishizone, Tomoyuki Higuchi · 2024

Read two stages domain invariant representation learners solve the large co-variate shift in unsupervised domain adaptation with two dimensional data domains by Hisashi Oshima, Tsuyoshi Ishizone, Tomoyuki Higuchi (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-04682.txt.

Description

Hisashi Oshima and co-authors proposed in 2024 a two-stage method for adapting a classifier to shifted new data with no labels. A known failure case is training on plain handwritten digits. Testing then uses colored street digits instead.

Their method learns features that stay steady across domains in two hops, through unlabeled middle data. Steady features are learned between source and middle, and between middle and target, at the same time. A theorem measures the gap to the unknown target rules, guiding free settings. Across thirty-eight tasks on four data sets, the method beat past approaches.

Catalog record

CREATORHisashi Oshima, Tsuyoshi Ishizone, Tomoyuki Higuchi
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE64 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-04682

Browse the full library · About the Tower of Babel