Reproduce “Simple Meta-Learning for HPO and CASH Using Performance Representations” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Emiel Witting, Cheng Yan, Tom Julian Viering
The paper introduces Linear-Perf, a Bayesian ridge regression surrogate for combined algorithm selection and hyperparameter optimization (CASH). It represents each candidate by its performance across prior tasks, rather than its hyperparameter values. On fixed TabArena CASH candidate sets, Linear-Perf reaches ABLR’s regret level in 73 evaluations, compared with 149 for ABLR; this measures benchmark evaluations, not elapsed training time.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.