MASIF: Meta-learned Algorithm Selection using Implicit Fidelity Information

Published: 15 Aug 2023, Last Modified: 15 Aug 2023AutoML 2023 (Journal Track)EveryoneRevisionsBibTeX
Paper URL:
Broader Impact Statement: pdf
Paper Availability And License: Yes
Code Of Conduct: Yes
TLDR: Bridging Multi-Fidelity and Algorithm Selection using Meta-Learning on Learning Curves. Use a transformer to jointly interpret curves from the candidate algorithms of varying lengths in a non-parametric and non-myopic way for Algorithm Selection.
Submission Number: 7