TransPL: VQ-Code Transition Matrices for Pseudo-Labeling of Time Series Unsupervised Domain Adaptation

Jaeho Kim, Seulki Lee · 2025

Read TransPL: VQ-Code Transition Matrices for Pseudo-Labeling of Time Series Unsupervised Domain Adaptation by Jaeho Kim, Seulki Lee (2025) 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-2505-09955.txt.

Description

Jaeho Kim and Seulki Lee improve how machines guess labels when a time series moves to a new domain. Their 2025 paper tackles unsupervised domain adaptation: a model trained on labeled data must work where labels are absent and habits differ. Standard guessing tricks miss the rhythm of sequences and the quirks of each sensor channel.

Their method, TransPL, first compresses source sequences into codes, then builds matrices of how codes follow each other, per class and per channel. Bayes' rule turns those source habits into target guesses, weighted channel by channel. On four benchmark suites the method beat the best prior guessing schemes by clear margins in accuracy and balance. Its guesses also come with a readable trail, which the authors count as a third virtue beside accuracy.

Catalog record

CREATORJaeho Kim, Seulki Lee
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE62 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-09955

Browse the full library · About the Tower of Babel