Reproduce “Mixed Variable Bayesian Optimization with Frequency Modulated Kernels” 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
Changyong Oh, Efstratios Gavves, Max Welling
The authors introduce a Gaussian-process kernel for mixed-variable Bayesian optimization. It uses distances between continuous inputs to modulate graph-Fourier components representing discrete choices, allowing the variable types to interact. The kernel has conditions for valid covariance and similarity ordering. In architecture-and-training-setting search, MODLAP reached an objective score of 0.06850 after 200 evaluations, versus RE and BOHB at 600, on this benchmark.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.