Are encoders able to learn landmarkers for warm-starting of Hyperparameter Optimization?

Antoni Zajko, Katarzyna Woźnica · 2025

Read Are encoders able to learn landmarkers for warm-starting of Hyperparameter Optimization? by Antoni Zajko, Katarzyna Woźnica (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-2507-12604.txt.

Description

Antoni Zajko and Katarzyna Woźnica test whether encoders can learn landmarkers for warm-starting hyperparameter optimization, in their 2025 arXiv paper. Meta-learning needs good representations of tabular datasets, and universal ones are the usual aim. Their target is narrower: representations should capture landmarker properties, meaning cheap probe models whose performance hints at dataset character.

They propose two methods, one based on deep metric learning and one on reconstructing landmarkers. Both were judged on the warm-starting task itself and on how well the representations met the landmarker requirement. The encoders did learn landmarker-aligned representations, but that did not translate into large gains on the optimization task, a nuanced result for meta-learning design.

Catalog record

CREATORAntoni Zajko, Katarzyna Woźnica
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE30 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-12604

Browse the full library · About the Tower of Babel